QML_HEP_GSoC_2021_Tasks_IV_GNN.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "QML-HEP GSoC 2021 Tasks IV GNN.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "GTRtFvj-jhmn"
      },
      "source": [
        "import numpy as np\n",
        "import tensorflow as tf\n",
        "from tensorflow import keras\n",
        "\n",
        "# visualization tools\n",
        "%matplotlib inline\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "I9nD1ES_u8Uw"
      },
      "source": [
        "import logging\n",
        "logging.basicConfig(level=logging.INFO, format='[%(asctime)s] %(levelname)s: %(message)s')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4Ub_EjJnrg7N",
        "outputId": "850cb5f9-be65-49ac-9827-5c4d2b73321e"
      },
      "source": [
        "# Mount the Google Drive as folder directory for the notebook\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Yo2O1exsHTXG",
        "outputId": "69b1140c-59cf-4299-b368-4fedf9f1ee41"
      },
      "source": [
        "%cd '/content/drive/MyDrive/Projects/GSoC 2021'"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/content/drive/MyDrive/Projects/GSoC 2021\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "g-jxhzu1oEHG"
      },
      "source": [
        "# Note: please change the current directory to the correct directory of the .py file if you want to run this code\n",
        "from tf_keras_model import get_particle_net, get_particle_net_lite\n",
        "from tf_keras_model_static import get_particle_net_static, get_particle_net_lite_static"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ewn3h8qtAnCF"
      },
      "source": [
        "# Task IV: Classical Graph Neural Network (GNN) Part\n",
        "\n",
        "Problem statement: use ParticleNet’s data for Quark/Gluon jet classification[1], then choose 2 Graph-based architectures of your choice to classify jets as being quarks (label = 1) or gluons (label = 0). Provide a description on what considerations you have taken to project this point-cloud dataset to a set of interconnected nodes and edges. Discuss the resulting performance of the 2 chosen architectures."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vw-OTcxJzFfu"
      },
      "source": [
        "## The Approach Used to Tackle the Problem"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "n0ZPewoTQqri"
      },
      "source": [
        "### Dataset and Hardware Limitation"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pBwaPvc_MtL_"
      },
      "source": [
        "Since Google Colab offers a limited ammount of RAM, this project only used the first 9 files of data (without charm and bottom jets) from reference [1]. This is also need to be done considering limited ammount of time available for the training process. Each file has 100k jets, exactly 50k quark and 50k gluon jets, randomly sorted (label imbalance is not an issue). Every jet in the file contains $M$ x $F$ data where $M$ is the maximum number of multiplicity of the jets in the file and $F$ is the number of each particle's features (pt, rapidity, azimuthal angle, and pdgid/Particle Data Group ID). So, in total, 900k jets are used in this project."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AoOZCP4AQugv"
      },
      "source": [
        "### Point-cloud - to - Graph Projection and Choosing Graph-based Architectures"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7cxDRn1TPx5g"
      },
      "source": [
        "How to project the point-cloud dataset to graphs is strongly interdependent with the network architectures. This project will follow mainly the reference [2] but with some changes to explore things that is not yet explored in the paper."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JhfM5BPdT4xA"
      },
      "source": [
        "ParticleNet, a state-of-the-art graph-based network architecture recently published, proposed to treat a jet as an unordered set of its constituent particles or \"particle cloud\"[2]. One main advantage of using this representation is that incorporating additional nonadditive information of the particles, like the pdgid, is straightforward. It is proven in reference [2] that including the pdgid data in the model's training improved the model's performance."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "udStPkElR7ao"
      },
      "source": [
        "The paper utilized Dynamic Graph Convolutional Neural Network (DGCNN) with \"EdgeConv\" operation as its unique selling point[3]. This operation treats every particle in the particle cloud as a vertice in a graph with its k nearest neighboring particles as the edges, explicitly exploits the local spatial structure of particle clouds."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Kz2AzM0YNMxq"
      },
      "source": [
        "![edgeconv block.jpg](data:image/jpeg;base64,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)\n",
        "\n",
        "**Figure 1**: The inner structure of an EdgeConv block[2]."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GSmc7gD3T1na"
      },
      "source": [
        "Equation below defines an EdgeConv operation on a point $x_{i}$ used in reference [2]\n",
        "$$\n",
        "\\boldsymbol{x}_{i}^{\\prime}=\\frac{1}{k}\\sum_{j=1}^{k} h_{\\Theta}\\left(\\boldsymbol{x}_{i}, \\boldsymbol{x}_{i_{j}}-\\boldsymbol{x}_{i}\\right)\n",
        "$$\n",
        "where $\\boldsymbol{x}_{i}$ is the feature vector of the point $x_{i}$, $\\boldsymbol{x}_{i_{j}}$ denoted the feature vector of the k nearest neighboring points of point $x_{i}$ with index $j$, and $h_{\\Theta}$ is a multilayer perceptron whose parameters are shared between all edges. One time of this operation application is depicted as an \"EdgeConv block\" in the **Figure 2**.\n",
        "\n",
        "Reference [2] also explain how they update the graph \"dynamically\" along the network. The set of k nearest neighboring particles for every particle is updated at the beginning of every EdgeConv block. The first block will use the rapidity and azimuthal angle to determine the distances, while the rest of the block will use the learned feature vectors from the previous block as coordinates."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5loOEcgQOgyG"
      },
      "source": [
        "![particlenet-lite.jpg](data:image/jpeg;base64,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)\n",
        "\n",
        "**Figure 2**: ParticleNet-Lite networks architecture[2]. $k$ is the number of nearest neighboring particles, and $C$ = ($C_{1}$, $C_{2}$, $C_{3}$) is the number of channels."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Drucup7kT6F7"
      },
      "source": [
        "After reading reference [2] and its [code implementation](https://github.com/hqucms/ParticleNet), there are three important parts which in my opinion are keys in obtaining the state-of-the-art performance achieved by the paper:\n",
        "1. The EdgeConv operation itself that exploits the local spatial structure of particle clouds.\n",
        "2. The \"dynamic\" aspect of updating the graph by changing the set of k nearest neighboring particles for every particle along the network.\n",
        "3. The particle cloud representation."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RM6SJEQoZoT9"
      },
      "source": [
        "I think it is interesting to explore which of them gives more significant improvement to the performance since the original paper does not discuss about this. Therefore, this project will follow the ParticleNet architecture implemented in reference [2] (the \"lite\" version will be used to shorten the training time needed) and compare it with the modified version that doesn't dynamically update the graph. Every EdgeConv block in this modified version will use the same set of nearest neighboring particles as the first block, which is to use the rapidity and azimuthal angle data in the dataset.\n",
        "\n",
        "By comparing the performance of these two architectures, we should be able to conclude the importance of the \"dynamic\" aspect in ParticleNet architecture, at least for the \"lite\" version."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GyoQJqN0PvpG"
      },
      "source": [
        "## Code Implementation"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kxjM01RLQ3ax"
      },
      "source": [
        "In this project, all features in the dataset (pt, rapidity, azimuthal angle and pdgid) will be used as the particle's features. The rapidity and azimuthal angle will be used to calculate the distance for k-NN."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ke-o_bYfSQL8"
      },
      "source": [
        "Since each file have different value of $M$, all jets in every file will be padded with zeros to have the same size of $M$. The particles will then be masked with a \"mask\" array that gives information whether the particle is a real particle from the dataset or just a padded \"zero-particles\"."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_zOVO3pojKyz"
      },
      "source": [
        "### Load and Check the Dataset"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "RVB6yZP5Tqxa"
      },
      "source": [
        "The function below will do the $M$ padding between two dataset files."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0s87xZoT8Rv6"
      },
      "source": [
        "def dataset_padding(X_1, X_2, y_1, y_2):\n",
        "    if X_1.shape[1] != X_2.shape[1]:\n",
        "        if X_1.shape[1] > X_2.shape[1]:\n",
        "            X_2 = np.concatenate((X_2,np.zeros((X_2.shape[0], X_1.shape[1]-X_2.shape[1], X_2.shape[2]))), axis=1)\n",
        "        else:\n",
        "            X_1 = np.concatenate((X_1,np.zeros((X_1.shape[0], X_2.shape[1]-X_1.shape[1], X_1.shape[2]))), axis=1)\n",
        "\n",
        "        X_1 = np.concatenate((X_1,X_2), axis=0)\n",
        "        y_1 = np.concatenate((y_1,y_2), axis=0)\n",
        "\n",
        "    return X_1, y_1"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fea6xmguT4_l"
      },
      "source": [
        "Let's load the first 9 files of the dataset from reference [1]. Since there are two classes to be classified, quarks (label = 1) or gluons (label = 0), and a softmax activation function is used at the end of the ParticleNet-Lite networks, the labels need to be transformed to one-hot format."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AhEgXDn6ApSN",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "38784752-3bb4-4c38-e97b-856729b12e77"
      },
      "source": [
        "# Load the dataset\n",
        "# Note: please change the directory to the correct directory of the dataset if you want to run this code\n",
        "with np.load('./jet-data/QG_jets.npz') as data:\n",
        "    X_train = data[\"X\"]\n",
        "    y_train = data[\"y\"]\n",
        "    print(\"Dataset QG_jets.npz has been loaded.\")\n",
        "    \n",
        "for i in range(8):\n",
        "    with np.load('./jet-data/QG_jets_' + str(i+1) + '.npz') as data:\n",
        "        temp_X = data[\"X\"]\n",
        "        temp_y = data[\"y\"]\n",
        "        print(\"Dataset QG_jets_\" + str(i+1) + \".npz has been loaded.\")\n",
        "        X_train, y_train = dataset_padding(X_train, temp_X, y_train, temp_y)\n",
        "\n",
        "# Transform the labels to one-hot format\n",
        "y_train = tf.keras.utils.to_categorical(y_train)\n",
        "\n",
        "# Sanity check\n",
        "print(\"Whole dataset shape:\", X_train.shape, y_train.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Dataset QG_jets.npz has been loaded.\n",
            "Dataset QG_jets_1.npz has been loaded.\n",
            "Dataset QG_jets_2.npz has been loaded.\n",
            "Dataset QG_jets_3.npz has been loaded.\n",
            "Dataset QG_jets_4.npz has been loaded.\n",
            "Dataset QG_jets_5.npz has been loaded.\n",
            "Dataset QG_jets_6.npz has been loaded.\n",
            "Dataset QG_jets_7.npz has been loaded.\n",
            "Dataset QG_jets_8.npz has been loaded.\n",
            "Whole dataset shape: (900000, 142, 4) (900000, 2)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p6wbuQg4qYW1",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d884803d-05e5-456a-ac89-1d82d06d18ff"
      },
      "source": [
        "num_samples = X_train.shape[0]\n",
        "num_particles = X_train.shape[1]\n",
        "num_features = X_train.shape[2]\n",
        "\n",
        "print(\"Total number of samples:\", num_samples)\n",
        "print(\"Maximum number of particles in a jet:\", num_particles)\n",
        "print(\"Number of features:\", num_features)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Total number of samples: 900000\n",
            "Maximum number of particles in a jet: 142\n",
            "Number of features: 4\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "l5f1NOpdEAGx"
      },
      "source": [
        "### Preprocess the Dataset"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0VFMXX7UWAGG"
      },
      "source": [
        "Split 80% of the data for training set, 10% for validation set, and 10% for testing set as suggested in the reference [2]."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3dqSXXerBSaF",
        "outputId": "2aafa6f3-292b-470e-8cf9-4aaba1a4174d"
      },
      "source": [
        "from sklearn.utils import shuffle\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "X_train, y_train = shuffle(X_train, y_train, random_state=2021)\n",
        "X_train, X_val, y_train, y_val = train_test_split(X_train, y_train, test_size=0.2, random_state=2021)\n",
        "X_val, X_test, y_val, y_test = train_test_split(X_val, y_val, test_size=0.5, random_state=2021)\n",
        "\n",
        "# Sanity check\n",
        "print(\"Train dataset shape:\", X_train.shape, y_train.shape)\n",
        "print(\"Val dataset shape:\", X_val.shape, y_val.shape)\n",
        "print(\"Test dataset shape:\", X_test.shape, y_test.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train dataset shape: (720000, 142, 4) (720000, 2)\n",
            "Val dataset shape: (90000, 142, 4) (90000, 2)\n",
            "Test dataset shape: (90000, 142, 4) (90000, 2)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9Dx5KFTXWR4t"
      },
      "source": [
        "Create a dictionary that contains the \"points\", \"features\", and \"mask\" array for training set, validation set, and testing set.\n",
        "\n",
        "- \"points\" contains only rapidity and azimuthal angle value, this will be used to calculate distance for k-NN.\n",
        "- \"features\" contains all the features (pt, rapidity, azimuthal angle), this will be used as the particle's features for training.\n",
        "- \"mask\" gives information whether the particle is a real particle from the dataset or just a padded \"zero-particles\" (taking value of zero for the padded positions)."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Rhx9_Uywpt2o"
      },
      "source": [
        "train_dict = {\n",
        "    # the coordinates of the particles in the (eta, phi) space (shape=num_samples,num_particles,2)\n",
        "    'points': X_train[:, :, 1:3],\n",
        "    # the features of the particles (shape=num_samples,num_particles,num_features)\n",
        "    'features': X_train,\n",
        "    # a mask array, taking a value of 0 for padded positions (shape=num_samples,num_particles,1)\n",
        "    'mask': np.sum(X_train, axis=2).reshape(len(X_train), num_particles, 1)\n",
        "}\n",
        "\n",
        "val_dict = {\n",
        "    # the coordinates of the particles in the (eta, phi) space (shape=num_samples,num_particles,2)\n",
        "    'points': X_val[:, :, 1:3],\n",
        "    # the features of the particles (shape=num_samples,num_particles,num_features)\n",
        "    'features': X_val,\n",
        "    # a mask array, taking a value of 0 for padded positions (shape=num_samples,num_particles,1)\n",
        "    'mask': np.sum(X_val, axis=2).reshape(len(X_val), num_particles, 1)\n",
        "}\n",
        "\n",
        "test_dict = {\n",
        "    # the coordinates of the particles in the (eta, phi) space (shape=num_samples,num_particles,2)\n",
        "    'points': X_test[:, :, 1:3],\n",
        "    # the features of the particles (shape=num_samples,num_particles,num_features)\n",
        "    'features': X_test,\n",
        "    # a mask array, taking a value of 0 for padded positions (shape=num_samples,num_particles,1)\n",
        "    'mask': np.sum(X_test, axis=2).reshape(len(X_test), num_particles, 1)\n",
        "}"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VDa-l0dWItYX"
      },
      "source": [
        "### Dynamic Graph CNN"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "klu1giYuXzW9"
      },
      "source": [
        "Now let's start first with the \"dynamic\" version, the original architecture proposed in reference [2]."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2CRE93YxIgRh"
      },
      "source": [
        "#### Initialize and Train the Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VSUNNj5uj7uv"
      },
      "source": [
        "Initialize the dynamic/original version of the ParticleNet-Lite model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ht28Rpr_tYm6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "312e19b8-ed1d-4b0d-e5eb-ada7165f0069"
      },
      "source": [
        "model_type = 'particle_net_lite'\n",
        "num_classes = 2\n",
        "input_shapes = {k:train_dict[k].shape[1:] for k in train_dict}\n",
        "\n",
        "if 'lite' in model_type:\n",
        "    model_dynamic = get_particle_net_lite(num_classes, input_shapes)\n",
        "else:\n",
        "    model_dynamic = get_particle_net(num_classes, input_shapes)\n",
        "\n",
        "print(\"Input shapes:\", input_shapes)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Input shapes: {'points': (142, 2), 'features': (142, 4), 'mask': (142, 1)}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CuEcC1JvYNjH"
      },
      "source": [
        "The hyperparameters ($k$, $C$, batch size, and learning rate) used in this project will be kept the same as the original paper so the comparison will be fair, both for the \"dynamic\" and \"static\" version.\n",
        "\n",
        "The number of epochs is chosen as 20 based on trial and error. I found that the training and validation performance start to converge after about 15 epochs."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LkLIStO9uksK"
      },
      "source": [
        "# Training parameters\n",
        "batch_size = 1024 if 'lite' in model_type else 384\n",
        "epochs = 20"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Xghu-aWwuw74"
      },
      "source": [
        "def lr_schedule(epoch):\n",
        "    lr = 1e-3\n",
        "    if epoch > 10:\n",
        "        lr *= 0.1\n",
        "    elif epoch > 20:\n",
        "        lr *= 0.01\n",
        "    logging.info('Learning rate: %f'%lr)\n",
        "    return lr"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0tlwVmKBuzCn",
        "outputId": "52d761fd-eb69-4d89-bd35-0192d86d6940"
      },
      "source": [
        "model_dynamic.compile(loss='categorical_crossentropy',\n",
        "              optimizer=keras.optimizers.Adam(learning_rate=lr_schedule(0)),\n",
        "              metrics=['accuracy', tf.keras.metrics.AUC()])\n",
        "model_dynamic.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:50:31,173] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Model: \"ParticleNet_Lite_Dynamic\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "mask (InputLayer)               [(None, 142, 1)]     0                                            \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.not_equal_1 (TFOpLambda (None, 142, 1)       0           mask[0][0]                       \n",
            "__________________________________________________________________________________________________\n",
            "tf.cast_2 (TFOpLambda)          (None, 142, 1)       0           tf.math.not_equal_1[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.equal_1 (TFOpLambda)    (None, 142, 1)       0           tf.cast_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.cast_3 (TFOpLambda)          (None, 142, 1)       0           tf.math.equal_1[0][0]            \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_5 (TFOpLambda) (None, 142, 1)       0           tf.cast_3[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "points (InputLayer)             [(None, 142, 2)]     0                                            \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.add_2 (TFOpLambda)      (None, 142, 2)       0           tf.math.multiply_5[0][0]         \n",
            "                                                                 points[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "features (InputLayer)           [(None, 142, 4)]     0                                            \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.transpose_2 (TFOpL (None, 2, 142)       0           tf.math.add_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_7 (TFOpLambda)   (None, 142, 1, 4)    0           features[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_6 (TFOpLambda) (None, 142, 2)       0           tf.math.add_2[0][0]              \n",
            "                                                                 tf.math.add_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "tf.linalg.matmul_1 (TFOpLambda) (None, 142, 142)     0           tf.math.add_2[0][0]              \n",
            "                                                                 tf.compat.v1.transpose_2[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_7 (TFOpLambda) (None, 142, 2)       0           tf.math.add_2[0][0]              \n",
            "                                                                 tf.math.add_2[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_fts_bn (BatchNormal (None, 142, 1, 4)    16          tf.expand_dims_7[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_sum_2 (TFOpLambd (None, 142, 1)       0           tf.math.multiply_6[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_8 (TFOpLambda) (None, 142, 142)     0           tf.linalg.matmul_1[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_sum_3 (TFOpLambd (None, 142, 1)       0           tf.math.multiply_7[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.squeeze_3 (TFOpLam (None, 142, 4)       0           ParticleNet_fts_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.subtract_3 (TFOpLambda) (None, 142, 142)     0           tf.math.reduce_sum_2[0][0]       \n",
            "                                                                 tf.math.multiply_8[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.transpose_3 (TFOpL (None, 1, 142)       0           tf.math.reduce_sum_3[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.shape_2 (TFOpLambd (3,)                 0           tf.compat.v1.squeeze_3[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_3 (TFOpLam (None, 142, 142)     0           tf.math.subtract_3[0][0]         \n",
            "                                                                 tf.compat.v1.transpose_3[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.getitem_4 (Sli ()                   0           tf.compat.v1.shape_2[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.negative_1 (TFOpLambda) (None, 142, 142)     0           tf.__operators__.add_3[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.range_2 (TFOpLambda)         (None,)              0           tf.__operators__.getitem_4[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.top_k_1 (TFOpLambda)    TopKV2(values=(None, 0           tf.math.negative_1[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.reshape_2 (TFOpLambda)       (None, 1, 1, 1)      0           tf.range_2[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.getitem_3 (Sli (None, 142, 7)       0           tf.math.top_k_1[0][1]            \n",
            "__________________________________________________________________________________________________\n",
            "tf.tile_4 (TFOpLambda)          (None, 142, 7, 1)    0           tf.reshape_2[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_8 (TFOpLambda)   (None, 142, 7, 1)    0           tf.__operators__.getitem_3[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_9 (TFOpLambda)   (None, 142, 1, 4)    0           tf.compat.v1.squeeze_3[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.concat_4 (TFOpLambda)        (None, 142, 7, 2)    0           tf.tile_4[0][0]                  \n",
            "                                                                 tf.expand_dims_8[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "tf.tile_5 (TFOpLambda)          (None, 142, 7, 4)    0           tf.expand_dims_9[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.gather_nd_2 (TFOpL (None, 142, 7, 4)    0           tf.compat.v1.squeeze_3[0][0]     \n",
            "                                                                 tf.concat_4[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.subtract_4 (TFOpLambda) (None, 142, 7, 4)    0           tf.compat.v1.gather_nd_2[0][0]   \n",
            "                                                                 tf.tile_5[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.concat_5 (TFOpLambda)        (None, 142, 7, 8)    0           tf.tile_5[0][0]                  \n",
            "                                                                 tf.math.subtract_4[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_conv0 (Co (None, 142, 7, 32)   256         tf.concat_5[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_bn0 (Batc (None, 142, 7, 32)   128         ParticleNet_EdgeConv0_conv0[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_act0 (Act (None, 142, 7, 32)   0           ParticleNet_EdgeConv0_bn0[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_conv1 (Co (None, 142, 7, 32)   1024        ParticleNet_EdgeConv0_act0[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_bn1 (Batc (None, 142, 7, 32)   128         ParticleNet_EdgeConv0_conv1[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_act1 (Act (None, 142, 7, 32)   0           ParticleNet_EdgeConv0_bn1[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_10 (TFOpLambda)  (None, 142, 1, 4)    0           tf.compat.v1.squeeze_3[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_conv2 (Co (None, 142, 7, 32)   1024        ParticleNet_EdgeConv0_act1[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_sc_conv ( (None, 142, 1, 32)   128         tf.expand_dims_10[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_bn2 (Batc (None, 142, 7, 32)   128         ParticleNet_EdgeConv0_conv2[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_sc_bn (Ba (None, 142, 1, 32)   128         ParticleNet_EdgeConv0_sc_conv[0][\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_act2 (Act (None, 142, 7, 32)   0           ParticleNet_EdgeConv0_bn2[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.squeeze_4 (TFOpLam (None, 142, 32)      0           ParticleNet_EdgeConv0_sc_bn[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_3 (TFOpLamb (None, 142, 32)      0           ParticleNet_EdgeConv0_act2[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_4 (TFOpLam (None, 142, 32)      0           tf.compat.v1.squeeze_4[0][0]     \n",
            "                                                                 tf.math.reduce_mean_3[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_sc_act (A (None, 142, 32)      0           tf.__operators__.add_4[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.add_3 (TFOpLambda)      (None, 142, 32)      0           tf.math.multiply_5[0][0]         \n",
            "                                                                 ParticleNet_EdgeConv0_sc_act[0][0\n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.transpose_4 (TFOpL (None, 32, 142)      0           tf.math.add_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_9 (TFOpLambda) (None, 142, 32)      0           tf.math.add_3[0][0]              \n",
            "                                                                 tf.math.add_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "tf.linalg.matmul_2 (TFOpLambda) (None, 142, 142)     0           tf.math.add_3[0][0]              \n",
            "                                                                 tf.compat.v1.transpose_4[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_10 (TFOpLambda (None, 142, 32)      0           tf.math.add_3[0][0]              \n",
            "                                                                 tf.math.add_3[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_sum_4 (TFOpLambd (None, 142, 1)       0           tf.math.multiply_9[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_11 (TFOpLambda (None, 142, 142)     0           tf.linalg.matmul_2[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_sum_5 (TFOpLambd (None, 142, 1)       0           tf.math.multiply_10[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.subtract_5 (TFOpLambda) (None, 142, 142)     0           tf.math.reduce_sum_4[0][0]       \n",
            "                                                                 tf.math.multiply_11[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.transpose_5 (TFOpL (None, 1, 142)       0           tf.math.reduce_sum_5[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.shape_3 (TFOpLambd (3,)                 0           ParticleNet_EdgeConv0_sc_act[0][0\n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_5 (TFOpLam (None, 142, 142)     0           tf.math.subtract_5[0][0]         \n",
            "                                                                 tf.compat.v1.transpose_5[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.getitem_6 (Sli ()                   0           tf.compat.v1.shape_3[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.negative_2 (TFOpLambda) (None, 142, 142)     0           tf.__operators__.add_5[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.range_3 (TFOpLambda)         (None,)              0           tf.__operators__.getitem_6[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.top_k_2 (TFOpLambda)    TopKV2(values=(None, 0           tf.math.negative_2[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.reshape_3 (TFOpLambda)       (None, 1, 1, 1)      0           tf.range_3[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.getitem_5 (Sli (None, 142, 7)       0           tf.math.top_k_2[0][1]            \n",
            "__________________________________________________________________________________________________\n",
            "tf.tile_6 (TFOpLambda)          (None, 142, 7, 1)    0           tf.reshape_3[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_11 (TFOpLambda)  (None, 142, 7, 1)    0           tf.__operators__.getitem_5[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_12 (TFOpLambda)  (None, 142, 1, 32)   0           ParticleNet_EdgeConv0_sc_act[0][0\n",
            "__________________________________________________________________________________________________\n",
            "tf.concat_6 (TFOpLambda)        (None, 142, 7, 2)    0           tf.tile_6[0][0]                  \n",
            "                                                                 tf.expand_dims_11[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "tf.tile_7 (TFOpLambda)          (None, 142, 7, 32)   0           tf.expand_dims_12[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.gather_nd_3 (TFOpL (None, 142, 7, 32)   0           ParticleNet_EdgeConv0_sc_act[0][0\n",
            "                                                                 tf.concat_6[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.subtract_6 (TFOpLambda) (None, 142, 7, 32)   0           tf.compat.v1.gather_nd_3[0][0]   \n",
            "                                                                 tf.tile_7[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.concat_7 (TFOpLambda)        (None, 142, 7, 64)   0           tf.tile_7[0][0]                  \n",
            "                                                                 tf.math.subtract_6[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_conv0 (Co (None, 142, 7, 64)   4096        tf.concat_7[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_bn0 (Batc (None, 142, 7, 64)   256         ParticleNet_EdgeConv1_conv0[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_act0 (Act (None, 142, 7, 64)   0           ParticleNet_EdgeConv1_bn0[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_conv1 (Co (None, 142, 7, 64)   4096        ParticleNet_EdgeConv1_act0[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_bn1 (Batc (None, 142, 7, 64)   256         ParticleNet_EdgeConv1_conv1[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_act1 (Act (None, 142, 7, 64)   0           ParticleNet_EdgeConv1_bn1[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_13 (TFOpLambda)  (None, 142, 1, 32)   0           ParticleNet_EdgeConv0_sc_act[0][0\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_conv2 (Co (None, 142, 7, 64)   4096        ParticleNet_EdgeConv1_act1[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_sc_conv ( (None, 142, 1, 64)   2048        tf.expand_dims_13[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_bn2 (Batc (None, 142, 7, 64)   256         ParticleNet_EdgeConv1_conv2[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_sc_bn (Ba (None, 142, 1, 64)   256         ParticleNet_EdgeConv1_sc_conv[0][\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_act2 (Act (None, 142, 7, 64)   0           ParticleNet_EdgeConv1_bn2[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.squeeze_5 (TFOpLam (None, 142, 64)      0           ParticleNet_EdgeConv1_sc_bn[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_4 (TFOpLamb (None, 142, 64)      0           ParticleNet_EdgeConv1_act2[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_6 (TFOpLam (None, 142, 64)      0           tf.compat.v1.squeeze_5[0][0]     \n",
            "                                                                 tf.math.reduce_mean_4[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_sc_act (A (None, 142, 64)      0           tf.__operators__.add_6[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_12 (TFOpLambda (None, 142, 64)      0           ParticleNet_EdgeConv1_sc_act[0][0\n",
            "                                                                 tf.cast_2[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_5 (TFOpLamb (None, 64)           0           tf.math.multiply_12[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "dense_2 (Dense)                 (None, 128)          8320        tf.math.reduce_mean_5[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "dropout_1 (Dropout)             (None, 128)          0           dense_2[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dense_3 (Dense)                 (None, 2)            258         dropout_1[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 26,898\n",
            "Trainable params: 26,122\n",
            "Non-trainable params: 776\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G04Gbehgu3Ni"
      },
      "source": [
        "lr_scheduler = keras.callbacks.LearningRateScheduler(lr_schedule)\n",
        "progress_bar = keras.callbacks.ProgbarLogger()\n",
        "callbacks = [lr_scheduler, progress_bar]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oBXgOvyzZLvR"
      },
      "source": [
        "Let's train the model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zeFdDHsxvXGN",
        "outputId": "4413fce3-2002-4b3e-aba3-78365abb35bd"
      },
      "source": [
        "history_dgcnn = model_dynamic.fit(train_dict, y_train,\n",
        "          batch_size=batch_size,\n",
        "          epochs=epochs,\n",
        "          validation_data=(val_dict, y_val),\n",
        "          shuffle=True,\n",
        "          callbacks=callbacks)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:51:05,137] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "704/704 [==============================] - 316s 449ms/sample - loss: 0.4792 - accuracy: 0.7773 - auc_1: 0.8512 - val_loss: 0.4369 - val_accuracy: 0.8062 - val_auc_1: 0.8795\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:56:21,263] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 2/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4384 - accuracy: 0.8045 - auc_1: 0.8787 - val_loss: 0.4578 - val_accuracy: 0.7900 - val_auc_1: 0.8676\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:01:32,468] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 3/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4286 - accuracy: 0.8098 - auc_1: 0.8846 - val_loss: 0.4352 - val_accuracy: 0.8074 - val_auc_1: 0.8833\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:06:43,541] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 4/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4237 - accuracy: 0.8124 - auc_1: 0.8874 - val_loss: 0.4152 - val_accuracy: 0.8180 - val_auc_1: 0.8923\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:11:54,555] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 5/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4192 - accuracy: 0.8150 - auc_1: 0.8900 - val_loss: 0.4231 - val_accuracy: 0.8163 - val_auc_1: 0.8914\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:17:05,453] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 6/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4179 - accuracy: 0.8158 - auc_1: 0.8907 - val_loss: 0.4101 - val_accuracy: 0.8215 - val_auc_1: 0.8962\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:22:16,460] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 7/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4146 - accuracy: 0.8173 - auc_1: 0.8925 - val_loss: 0.4169 - val_accuracy: 0.8157 - val_auc_1: 0.8916\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:27:27,432] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 8/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4145 - accuracy: 0.8175 - auc_1: 0.8926 - val_loss: 0.4096 - val_accuracy: 0.8194 - val_auc_1: 0.8953\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:32:38,470] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 9/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4117 - accuracy: 0.8187 - auc_1: 0.8942 - val_loss: 0.4068 - val_accuracy: 0.8227 - val_auc_1: 0.8971\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:37:49,609] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 10/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4112 - accuracy: 0.8191 - auc_1: 0.8944 - val_loss: 0.4071 - val_accuracy: 0.8233 - val_auc_1: 0.8964\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:43:00,825] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 11/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4101 - accuracy: 0.8194 - auc_1: 0.8950 - val_loss: 0.4156 - val_accuracy: 0.8159 - val_auc_1: 0.8921\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:48:11,837] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 12/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4052 - accuracy: 0.8222 - auc_1: 0.8976 - val_loss: 0.4011 - val_accuracy: 0.8246 - val_auc_1: 0.8998\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:53:23,080] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 13/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4046 - accuracy: 0.8228 - auc_1: 0.8980 - val_loss: 0.4002 - val_accuracy: 0.8257 - val_auc_1: 0.9004\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 14:58:34,360] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 14/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4041 - accuracy: 0.8231 - auc_1: 0.8982 - val_loss: 0.3990 - val_accuracy: 0.8265 - val_auc_1: 0.9010\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 15:03:45,529] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 15/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4038 - accuracy: 0.8229 - auc_1: 0.8984 - val_loss: 0.4008 - val_accuracy: 0.8249 - val_auc_1: 0.9000\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 15:08:56,784] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 16/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4043 - accuracy: 0.8230 - auc_1: 0.8981 - val_loss: 0.3987 - val_accuracy: 0.8264 - val_auc_1: 0.9010\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 15:14:07,933] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 17/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4034 - accuracy: 0.8232 - auc_1: 0.8986 - val_loss: 0.3987 - val_accuracy: 0.8262 - val_auc_1: 0.9012\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 15:19:19,088] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 18/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4038 - accuracy: 0.8228 - auc_1: 0.8984 - val_loss: 0.3984 - val_accuracy: 0.8265 - val_auc_1: 0.9012\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 15:24:30,250] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 19/20\n",
            "704/704 [==============================] - 311s 442ms/sample - loss: 0.4030 - accuracy: 0.8237 - auc_1: 0.8988 - val_loss: 0.3982 - val_accuracy: 0.8267 - val_auc_1: 0.9014\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 15:29:41,304] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 20/20\n",
            "704/704 [==============================] - 311s 441ms/sample - loss: 0.4034 - accuracy: 0.8234 - auc_1: 0.8986 - val_loss: 0.3983 - val_accuracy: 0.8266 - val_auc_1: 0.9012\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "o2v_brROaHS1"
      },
      "source": [
        "One important thing to be noticed here, as we will compare it later, is the time needed to train the model for 1 epoch, which is 311s/epoch."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AcPC1rznDd41"
      },
      "source": [
        "# save the training history as csv file\n",
        "import csv\n",
        "\n",
        "w = csv.writer(open(\"./train_history_dgcnn.csv\", \"w\"))\n",
        "for key, val in history_dgcnn.history.items():\n",
        "    w.writerow([key, val])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "IOlBlMDub33y",
        "outputId": "0296b3aa-9d26-4e4e-b73e-0e1ab0cab873"
      },
      "source": [
        "# save the last model\n",
        "model_dynamic.save(\"./model_save/model_dynamic\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: ./model_save/model_dynamic/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 15:39:11,821] INFO: Assets written to: ./model_save/model_dynamic/assets\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UYfeHLTKQV34"
      },
      "source": [
        "#### Plot the Training History"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "mIFOOjBQ28BU",
        "outputId": "ba01abc3-e10e-43ae-db29-863e561e30cb"
      },
      "source": [
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['auc_1'], color='blue', label='Training AUC')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['val_auc_1'], color='red', label='Validation AUC')\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"AUC\")\n",
        "plt.title(\"AUC of Dynamic Graph CNN Model\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9jpheR9Qvqz0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "48628746-4aa5-4f69-83a3-5297df9b8434"
      },
      "source": [
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['accuracy'], color='blue', label='Training Accuracy')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['val_accuracy'], color='red', label='Validation Accuracy')\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"Accuracy\")\n",
        "plt.title(\"Accuracy of Dynamic Graph CNN Model\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "LClrLPtCdfuF",
        "outputId": "82482e86-8f62-4f13-a4c9-1bc8f7a8a3b1"
      },
      "source": [
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['loss'], color='blue', label='Training Loss')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['val_loss'], color='red', label='Validation Loss')\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"Loss\")\n",
        "plt.title(\"Loss of Dynamic Graph CNN Model\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "65c937VIheKw"
      },
      "source": [
        "From the plots, it can be seen that the model converged before 20 epochs. The accuracy and AUC obtained are comparable to the values reported in reference [2], considering this project only used less than half of the available dataset (900k out of 2000k in total). A more in-depth discussion will be given later after the training of the \"static\" version of the model is done."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "f7HKFEzC3v7_"
      },
      "source": [
        "### Static Graph CNN"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yNlILFADi3va"
      },
      "source": [
        "For this static version, we will train the modified version of ParticleNet-Lite that does not update the graph along the EdgeConv operation."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ytPSwqA_HloO"
      },
      "source": [
        "#### Initialize and Train the Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WKmWHrj8kKFJ"
      },
      "source": [
        "Initialize the static/modified version of the ParticleNet-Lite model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CvymurtXHloP",
        "outputId": "f77e9cbc-0cc2-4c04-c5be-a014fe22906c"
      },
      "source": [
        "model_type = 'particle_net_lite_static'\n",
        "num_classes = 2\n",
        "input_shapes = {k:train_dict[k].shape[1:] for k in train_dict}\n",
        "\n",
        "if 'lite' in model_type:\n",
        "    model_static = get_particle_net_lite_static(num_classes, input_shapes)\n",
        "else:\n",
        "    model_static = get_particle_net_static(num_classes, input_shapes)\n",
        "\n",
        "print(\"Input shapes:\", input_shapes)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Input shapes: {'points': (142, 2), 'features': (142, 4), 'mask': (142, 1)}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "obwH5qBGjXwV"
      },
      "source": [
        "The exact same hyperparameters as the dynamic version will be used to ensure fairness."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GJc1ZWoMHloP"
      },
      "source": [
        "# Training parameters\n",
        "batch_size = 1024 if 'lite' in model_type else 384\n",
        "epochs = 20"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6dBayswCHloQ"
      },
      "source": [
        "def lr_schedule(epoch):\n",
        "    lr = 1e-3\n",
        "    if epoch > 10:\n",
        "        lr *= 0.1\n",
        "    elif epoch > 20:\n",
        "        lr *= 0.01\n",
        "    logging.info('Learning rate: %f'%lr)\n",
        "    return lr"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fZB2DqeDHloR",
        "outputId": "fe632e8d-41fa-430a-8de9-81ee0ab67e42"
      },
      "source": [
        "model_static.compile(loss='categorical_crossentropy',\n",
        "              optimizer=keras.optimizers.Adam(learning_rate=lr_schedule(0)),\n",
        "              metrics=['accuracy', tf.keras.metrics.AUC()])\n",
        "model_static.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:18:06,362] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Model: \"ParticleNet_Lite_Static\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "mask (InputLayer)               [(None, 142, 1)]     0                                            \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.not_equal (TFOpLambda)  (None, 142, 1)       0           mask[0][0]                       \n",
            "__________________________________________________________________________________________________\n",
            "tf.cast (TFOpLambda)            (None, 142, 1)       0           tf.math.not_equal[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.equal (TFOpLambda)      (None, 142, 1)       0           tf.cast[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "tf.cast_1 (TFOpLambda)          (None, 142, 1)       0           tf.math.equal[0][0]              \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply (TFOpLambda)   (None, 142, 1)       0           tf.cast_1[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "points (InputLayer)             [(None, 142, 2)]     0                                            \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.add (TFOpLambda)        (None, 142, 2)       0           tf.math.multiply[0][0]           \n",
            "                                                                 points[0][0]                     \n",
            "__________________________________________________________________________________________________\n",
            "features (InputLayer)           [(None, 142, 4)]     0                                            \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.transpose (TFOpLam (None, 2, 142)       0           tf.math.add[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims (TFOpLambda)     (None, 142, 1, 4)    0           features[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_1 (TFOpLambda) (None, 142, 2)       0           tf.math.add[0][0]                \n",
            "                                                                 tf.math.add[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "tf.linalg.matmul (TFOpLambda)   (None, 142, 142)     0           tf.math.add[0][0]                \n",
            "                                                                 tf.compat.v1.transpose[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_2 (TFOpLambda) (None, 142, 2)       0           tf.math.add[0][0]                \n",
            "                                                                 tf.math.add[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_fts_bn (BatchNormal (None, 142, 1, 4)    16          tf.expand_dims[0][0]             \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_sum (TFOpLambda) (None, 142, 1)       0           tf.math.multiply_1[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_3 (TFOpLambda) (None, 142, 142)     0           tf.linalg.matmul[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_sum_1 (TFOpLambd (None, 142, 1)       0           tf.math.multiply_2[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.squeeze (TFOpLambd (None, 142, 4)       0           ParticleNet_fts_bn[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.subtract (TFOpLambda)   (None, 142, 142)     0           tf.math.reduce_sum[0][0]         \n",
            "                                                                 tf.math.multiply_3[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.transpose_1 (TFOpL (None, 1, 142)       0           tf.math.reduce_sum_1[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.shape (TFOpLambda) (3,)                 0           tf.compat.v1.squeeze[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add (TFOpLambd (None, 142, 142)     0           tf.math.subtract[0][0]           \n",
            "                                                                 tf.compat.v1.transpose_1[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.getitem_1 (Sli ()                   0           tf.compat.v1.shape[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.negative (TFOpLambda)   (None, 142, 142)     0           tf.__operators__.add[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "tf.range (TFOpLambda)           (None,)              0           tf.__operators__.getitem_1[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.top_k (TFOpLambda)      TopKV2(values=(None, 0           tf.math.negative[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "tf.reshape (TFOpLambda)         (None, 1, 1, 1)      0           tf.range[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.getitem (Slici (None, 142, 7)       0           tf.math.top_k[0][1]              \n",
            "__________________________________________________________________________________________________\n",
            "tf.tile (TFOpLambda)            (None, 142, 7, 1)    0           tf.reshape[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_1 (TFOpLambda)   (None, 142, 7, 1)    0           tf.__operators__.getitem[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_2 (TFOpLambda)   (None, 142, 1, 4)    0           tf.compat.v1.squeeze[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "tf.concat (TFOpLambda)          (None, 142, 7, 2)    0           tf.tile[0][0]                    \n",
            "                                                                 tf.expand_dims_1[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "tf.tile_1 (TFOpLambda)          (None, 142, 7, 4)    0           tf.expand_dims_2[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.gather_nd (TFOpLam (None, 142, 7, 4)    0           tf.compat.v1.squeeze[0][0]       \n",
            "                                                                 tf.concat[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.subtract_1 (TFOpLambda) (None, 142, 7, 4)    0           tf.compat.v1.gather_nd[0][0]     \n",
            "                                                                 tf.tile_1[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.concat_1 (TFOpLambda)        (None, 142, 7, 8)    0           tf.tile_1[0][0]                  \n",
            "                                                                 tf.math.subtract_1[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_conv0 (Co (None, 142, 7, 32)   256         tf.concat_1[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_bn0 (Batc (None, 142, 7, 32)   128         ParticleNet_EdgeConv0_conv0[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_act0 (Act (None, 142, 7, 32)   0           ParticleNet_EdgeConv0_bn0[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_conv1 (Co (None, 142, 7, 32)   1024        ParticleNet_EdgeConv0_act0[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_bn1 (Batc (None, 142, 7, 32)   128         ParticleNet_EdgeConv0_conv1[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_act1 (Act (None, 142, 7, 32)   0           ParticleNet_EdgeConv0_bn1[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_3 (TFOpLambda)   (None, 142, 1, 4)    0           tf.compat.v1.squeeze[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_conv2 (Co (None, 142, 7, 32)   1024        ParticleNet_EdgeConv0_act1[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_sc_conv ( (None, 142, 1, 32)   128         tf.expand_dims_3[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_bn2 (Batc (None, 142, 7, 32)   128         ParticleNet_EdgeConv0_conv2[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_sc_bn (Ba (None, 142, 1, 32)   128         ParticleNet_EdgeConv0_sc_conv[0][\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_act2 (Act (None, 142, 7, 32)   0           ParticleNet_EdgeConv0_bn2[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.squeeze_1 (TFOpLam (None, 142, 32)      0           ParticleNet_EdgeConv0_sc_bn[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean (TFOpLambda (None, 142, 32)      0           ParticleNet_EdgeConv0_act2[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_1 (TFOpLam (None, 142, 32)      0           tf.compat.v1.squeeze_1[0][0]     \n",
            "                                                                 tf.math.reduce_mean[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv0_sc_act (A (None, 142, 32)      0           tf.__operators__.add_1[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.shape_1 (TFOpLambd (3,)                 0           ParticleNet_EdgeConv0_sc_act[0][0\n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.getitem_2 (Sli ()                   0           tf.compat.v1.shape_1[0][0]       \n",
            "__________________________________________________________________________________________________\n",
            "tf.range_1 (TFOpLambda)         (None,)              0           tf.__operators__.getitem_2[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.reshape_1 (TFOpLambda)       (None, 1, 1, 1)      0           tf.range_1[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "tf.tile_2 (TFOpLambda)          (None, 142, 7, 1)    0           tf.reshape_1[0][0]               \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_4 (TFOpLambda)   (None, 142, 7, 1)    0           tf.__operators__.getitem[0][0]   \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_5 (TFOpLambda)   (None, 142, 1, 32)   0           ParticleNet_EdgeConv0_sc_act[0][0\n",
            "__________________________________________________________________________________________________\n",
            "tf.concat_2 (TFOpLambda)        (None, 142, 7, 2)    0           tf.tile_2[0][0]                  \n",
            "                                                                 tf.expand_dims_4[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "tf.tile_3 (TFOpLambda)          (None, 142, 7, 32)   0           tf.expand_dims_5[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.gather_nd_1 (TFOpL (None, 142, 7, 32)   0           ParticleNet_EdgeConv0_sc_act[0][0\n",
            "                                                                 tf.concat_2[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.subtract_2 (TFOpLambda) (None, 142, 7, 32)   0           tf.compat.v1.gather_nd_1[0][0]   \n",
            "                                                                 tf.tile_3[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "tf.concat_3 (TFOpLambda)        (None, 142, 7, 64)   0           tf.tile_3[0][0]                  \n",
            "                                                                 tf.math.subtract_2[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_conv0 (Co (None, 142, 7, 64)   4096        tf.concat_3[0][0]                \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_bn0 (Batc (None, 142, 7, 64)   256         ParticleNet_EdgeConv1_conv0[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_act0 (Act (None, 142, 7, 64)   0           ParticleNet_EdgeConv1_bn0[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_conv1 (Co (None, 142, 7, 64)   4096        ParticleNet_EdgeConv1_act0[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_bn1 (Batc (None, 142, 7, 64)   256         ParticleNet_EdgeConv1_conv1[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_act1 (Act (None, 142, 7, 64)   0           ParticleNet_EdgeConv1_bn1[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "tf.expand_dims_6 (TFOpLambda)   (None, 142, 1, 32)   0           ParticleNet_EdgeConv0_sc_act[0][0\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_conv2 (Co (None, 142, 7, 64)   4096        ParticleNet_EdgeConv1_act1[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_sc_conv ( (None, 142, 1, 64)   2048        tf.expand_dims_6[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_bn2 (Batc (None, 142, 7, 64)   256         ParticleNet_EdgeConv1_conv2[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_sc_bn (Ba (None, 142, 1, 64)   256         ParticleNet_EdgeConv1_sc_conv[0][\n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_act2 (Act (None, 142, 7, 64)   0           ParticleNet_EdgeConv1_bn2[0][0]  \n",
            "__________________________________________________________________________________________________\n",
            "tf.compat.v1.squeeze_2 (TFOpLam (None, 142, 64)      0           ParticleNet_EdgeConv1_sc_bn[0][0]\n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_1 (TFOpLamb (None, 142, 64)      0           ParticleNet_EdgeConv1_act2[0][0] \n",
            "__________________________________________________________________________________________________\n",
            "tf.__operators__.add_2 (TFOpLam (None, 142, 64)      0           tf.compat.v1.squeeze_2[0][0]     \n",
            "                                                                 tf.math.reduce_mean_1[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "ParticleNet_EdgeConv1_sc_act (A (None, 142, 64)      0           tf.__operators__.add_2[0][0]     \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.multiply_4 (TFOpLambda) (None, 142, 64)      0           ParticleNet_EdgeConv1_sc_act[0][0\n",
            "                                                                 tf.cast[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "tf.math.reduce_mean_2 (TFOpLamb (None, 64)           0           tf.math.multiply_4[0][0]         \n",
            "__________________________________________________________________________________________________\n",
            "dense (Dense)                   (None, 128)          8320        tf.math.reduce_mean_2[0][0]      \n",
            "__________________________________________________________________________________________________\n",
            "dropout (Dropout)               (None, 128)          0           dense[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "dense_1 (Dense)                 (None, 2)            258         dropout[0][0]                    \n",
            "==================================================================================================\n",
            "Total params: 26,898\n",
            "Trainable params: 26,122\n",
            "Non-trainable params: 776\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dgeTBqWOHloS"
      },
      "source": [
        "lr_scheduler = keras.callbacks.LearningRateScheduler(lr_schedule)\n",
        "progress_bar = keras.callbacks.ProgbarLogger()\n",
        "callbacks = [lr_scheduler, progress_bar]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EUxEmy56kSRh"
      },
      "source": [
        "Let's train the model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8nOPddjDHloS",
        "outputId": "dc0a0a40-80cf-4edd-9d67-e320a68574dc"
      },
      "source": [
        "history_sgcnn = model_static.fit(train_dict, y_train,\n",
        "          batch_size=batch_size,\n",
        "          epochs=epochs,\n",
        "          validation_data=(val_dict, y_val),\n",
        "          shuffle=True,\n",
        "          callbacks=callbacks)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:18:30,277] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "704/704 [==============================] - 246s 349ms/sample - loss: 0.4681 - accuracy: 0.7867 - auc: 0.8595 - val_loss: 0.4412 - val_accuracy: 0.8037 - val_auc: 0.8778\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:22:35,878] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 2/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4306 - accuracy: 0.8093 - auc: 0.8834 - val_loss: 0.4394 - val_accuracy: 0.8044 - val_auc: 0.8796\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:26:07,350] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 3/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4234 - accuracy: 0.8124 - auc: 0.8876 - val_loss: 0.4166 - val_accuracy: 0.8173 - val_auc: 0.8916\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:29:38,794] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 4/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4185 - accuracy: 0.8153 - auc: 0.8904 - val_loss: 0.4133 - val_accuracy: 0.8190 - val_auc: 0.8932\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:33:10,159] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 5/20\n",
            "704/704 [==============================] - 212s 301ms/sample - loss: 0.4168 - accuracy: 0.8162 - auc: 0.8913 - val_loss: 0.4247 - val_accuracy: 0.8105 - val_auc: 0.8870\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:36:41,807] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 6/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4140 - accuracy: 0.8174 - auc: 0.8929 - val_loss: 0.4074 - val_accuracy: 0.8227 - val_auc: 0.8964\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:40:12,771] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 7/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4126 - accuracy: 0.8185 - auc: 0.8936 - val_loss: 0.4703 - val_accuracy: 0.7800 - val_auc: 0.8615\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:43:43,750] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 8/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4113 - accuracy: 0.8193 - auc: 0.8943 - val_loss: 0.4089 - val_accuracy: 0.8209 - val_auc: 0.8960\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:47:14,653] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 9/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4105 - accuracy: 0.8197 - auc: 0.8947 - val_loss: 0.4277 - val_accuracy: 0.8100 - val_auc: 0.8853\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:50:45,544] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 10/20\n",
            "704/704 [==============================] - 211s 299ms/sample - loss: 0.4094 - accuracy: 0.8203 - auc: 0.8954 - val_loss: 0.4284 - val_accuracy: 0.8108 - val_auc: 0.8870\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:54:16,353] INFO: Learning rate: 0.001000\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 11/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4078 - accuracy: 0.8213 - auc: 0.8962 - val_loss: 0.4121 - val_accuracy: 0.8202 - val_auc: 0.8941\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 12:57:47,333] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 12/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4030 - accuracy: 0.8230 - auc: 0.8987 - val_loss: 0.3966 - val_accuracy: 0.8288 - val_auc: 0.9021\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:01:18,341] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 13/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4033 - accuracy: 0.8232 - auc: 0.8986 - val_loss: 0.3969 - val_accuracy: 0.8281 - val_auc: 0.9019\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:04:49,385] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 14/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4024 - accuracy: 0.8237 - auc: 0.8991 - val_loss: 0.3958 - val_accuracy: 0.8289 - val_auc: 0.9024\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:08:20,350] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 15/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4015 - accuracy: 0.8237 - auc: 0.8996 - val_loss: 0.3965 - val_accuracy: 0.8291 - val_auc: 0.9022\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:11:51,505] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 16/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4018 - accuracy: 0.8239 - auc: 0.8994 - val_loss: 0.3957 - val_accuracy: 0.8287 - val_auc: 0.9025\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:15:22,778] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 17/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4021 - accuracy: 0.8242 - auc: 0.8992 - val_loss: 0.3956 - val_accuracy: 0.8287 - val_auc: 0.9026\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:18:53,765] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 18/20\n",
            "704/704 [==============================] - 211s 300ms/sample - loss: 0.4011 - accuracy: 0.8248 - auc: 0.8998 - val_loss: 0.3953 - val_accuracy: 0.8295 - val_auc: 0.9027\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:22:24,749] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 19/20\n",
            "704/704 [==============================] - 211s 299ms/sample - loss: 0.4015 - accuracy: 0.8247 - auc: 0.8996 - val_loss: 0.3950 - val_accuracy: 0.8294 - val_auc: 0.9030\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:25:55,560] INFO: Learning rate: 0.000100\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Epoch 20/20\n",
            "704/704 [==============================] - 211s 299ms/sample - loss: 0.4011 - accuracy: 0.8244 - auc: 0.8998 - val_loss: 0.3965 - val_accuracy: 0.8284 - val_auc: 0.9020\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PY7y5-lpkVxr"
      },
      "source": [
        "Same as before, notice that the time needed to train the static model is 211s/epoch. This will be compared later."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KrWrwkHdM-qq"
      },
      "source": [
        "# save the training history as csv file\n",
        "import csv\n",
        "\n",
        "w = csv.writer(open(\"./train_history_sgcnn.csv\", \"w\"))\n",
        "for key, val in history_sgcnn.history.items():\n",
        "    w.writerow([key, val])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "w-_Z0KJfA_ja",
        "outputId": "48b14b88-3c7a-4c24-edd8-1b90ed275e7c"
      },
      "source": [
        "# save the last model\n",
        "model_static.save(\"./model_save/model_static\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: ./model_save/model_static/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[2021-03-21 13:45:56,428] INFO: Assets written to: ./model_save/model_static/assets\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jM0uWHEQQmhe"
      },
      "source": [
        "#### Plot the Training History"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "iiZkkxpTHloU",
        "outputId": "98745a08-231e-4407-de93-aea3d433e087"
      },
      "source": [
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['auc'], color='blue', label='Training AUC')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['val_auc'], color='red', label='Validation AUC')\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"AUC\")\n",
        "plt.title(\"AUC of Static Graph CNN Model\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "QtS6ckknHloU",
        "outputId": "b0628155-fe95-469c-c6f8-460f42912373"
      },
      "source": [
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['accuracy'], color='blue', label='Training Accuracy')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['val_accuracy'], color='red', label='Validation Accuracy')\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"Accuracy\")\n",
        "plt.title(\"Accuracy of Static Graph CNN Model\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "rklpLYpDHloV",
        "outputId": "e3f20d5d-3f93-4ba5-9bf5-041bb8703a45"
      },
      "source": [
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['loss'], color='blue', label='Training Loss')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['val_loss'], color='red', label='Validation Loss')\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"Loss\")\n",
        "plt.title(\"Loss of Static Graph CNN Model\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_WE0c5r9kwpQ"
      },
      "source": [
        "One thing that really stood out from these plots is the instability of the training process. A more in-depth discussion will be given in the next section."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1FvF7ovWc6A_"
      },
      "source": [
        "### Comparing the Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "T3Ekp8zNnHdl"
      },
      "source": [
        "Let's put the results from both models side by side for easier comparison."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "rcBkAm3XebeS",
        "outputId": "337d6282-4460-4b4f-b3d8-1cc440f04ee9"
      },
      "source": [
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['auc_1'], color='teal', label='Training AUC - Dynamic')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['val_auc_1'], color='blue', label='Validation AUC - Dynamic')\n",
        "\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['auc'], color='orange', label='Training AUC - Static')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['val_auc'], color='red', label='Validation AUC - Static')\n",
        "\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"AUC\")\n",
        "plt.title(\"AUC of Dynamic vs Static Graph CNN Model\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "yx3xTN18fTxG",
        "outputId": "fdf0d610-c348-4834-c0b6-5937f44ba45a"
      },
      "source": [
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['accuracy'], color='teal', label='Training Accuracy - Dynamic')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['val_accuracy'], color='blue', label='Validation Accuracy - Dynamic')\n",
        "\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['accuracy'], color='orange', label='Training Accuracy - Static')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['val_accuracy'], color='red', label='Validation Accuracy - Static')\n",
        "\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"Accuracy\")\n",
        "plt.title(\"Accuracy of Dynamic vs Static Graph CNN Model\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "o8bhd7dxfoNr",
        "outputId": "a39d1c41-6734-4785-d25f-bc7ef4c2e0a1"
      },
      "source": [
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['loss'], color='teal', label='Training Loss - Dynamic')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_dgcnn.history['val_loss'], color='blue', label='Validation Loss - Dynamic')\n",
        "\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['loss'], color='orange', label='Training Loss - Static')\n",
        "plt.plot(np.linspace(1,epochs,epochs, dtype=int), history_sgcnn.history['val_loss'], color='red', label='Validation Loss - Static')\n",
        "\n",
        "plt.xlabel(\"Epochs\")\n",
        "plt.ylabel(\"Loss\")\n",
        "plt.title(\"Loss of Dynamic vs Static Graph CNN Model\")\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOydd3xUVfbAvyehBAidRFLQgNIJqSSIgEGMUjRIE7EG1oKrsLK6WNZVVuW3FlR017qsYgFBZWXRUAQkgiAlIEWaUiItYGghdJLc3x/3zTAJM8kkmckk4X4/n/nMvPvuve+8NzPvvHPPPeeKUgqDwWAwGIri52sBDAaDwVA5MQrCYDAYDE4xCsJgMBgMTjEKwmAwGAxOMQrCYDAYDE4xCsJgMBgMTjEK4hJDRF4QkUMicsDXspQXETkhIq18LYc3EJE7RORbX8vhDiKSJCJ7fS2HrxARJSJXuVGvyl0noyC8hIhkisj1vpbDERG5HHgU6KCUau5kf5KIFFg33hMisldEPheRLhUvbckopQKVUjt9cWwR6S4iy0UkR0SOiMgy23USkVQR+aEUfUVYN5katjKl1FSl1A1llC1eRL4RkaMickxENovIBBFpXJb+vIGI3CgiS0QkV0SyReR7EUmx9qVa12NckTZ7RSTJ+jzeqnOrw/4aVlmEi2OmW/ujipR/ZZUnefYsqz5GQVxaXA4cVkr9Xkyd/UqpQKA+0BXYCiwVkd4VIWBVQEQaAN8A/wSaAGHA34GzvpQLQES6AenAMqCdUqoR0AfIA6JctKnhrNxbiMgQ4AvgYyAcuAx4BrjZodoRYJyI1C+mqyPA30XEvxSH/wW420GWpsDVQHYp+rh0UEqZlxdeQCZwvZPy2sAkYL/1mgTUtvY1Q994jqF//EsBP2vf48A+IBfYBvR2cdyG6D9eNvAb8DT6QeB64DRQAJwApjhpmwTsdVL+LyDD+vwW8GqR/bOBsQ7n/RiwAcgBZgAB1r7G1vllA0etz+EO/aQDLwDLLRm/BpoCU4HjwGogwqG+Aq6yPtcBXrXOOQf4Aajj5Fy2ADc5bNew5IkFAoBPgcPWd7AauMxJH/HAMRfXvz1wBsi3zuGYVd4f+Mk6jz3AeIc2u61zOWG9rgZSgR8c6nQEFli/i4PAUy6O/wPwzxJ+m6loBfK6da4vAFcC31nbh6xr3qjI7/lJYLP13X3o8L0mAXvR1unvQBYwwsWxxTrfv5Qg3w/W9/+sQ/leIMn6PN6ScT1wj8N3qRx/I0X6TUcror2Av1X2MPBOkb5d/ket/X+xznE/MJLCv8PawETrHA8C72L9DnHx/6rML58LUF1fuFYQzwErgGAgCH0zfN7a9w/rB1XTevWw/lBtrZtKqFUvArjSxXE/Bv6HtgAi0E9Mf7D2FfsDdbUfuA6tWOoBCdYfw6a4mgGnsG6k1nmvAkLRT9dbgFHWvqbAYKCuJd8XwCyH46QD29E3q4bom9EvaOVWwzq3Dx3qO/4x37LahwH+QDfHP7VDm2eAqQ7b/YEt1ucH0DelulYfcUADJ300QN9IPwL6Ao2L7E/F4ebucG0j0cq6s3XzuMXh+1RADWd9WNcqC30DDrC2E53IVQ+tmJJK+G2moi2K0dZ1rQNcBSSjb3BBwBJgUpHf889AC+t7XQa84HBueejfdk2gn/WbaOzk2O2sc21Zgnw/ANFoZdTEKi+qID4FUoCd1nHdURD3At8Cfa2yVWiF7Nh3cf/RPtZ318m63tMo/Dt8Hf3A1MT6nr4G/uHO/68yvnwuQHV94VpB7AD6OWzfCGRan59D39yvKtLmKvST2fVAzWKO6Q+cQ/sYbGUPAOnW52J/oK72O/ypw6ztLUCy9flhYE6R877TYftl4F0Xx4sGjjpspwN/ddh+FZjrsH0zsM5hW1nXxg9tHUW58b1chbbC6lrbU4FnrM8jrZtBZzf6aQ9MsW4sedZNwaYkUymiIJy0nwS8bn2OoHgFMRz4yQ2Zwq1+2hW5/seAk8DTDn3vLqGvWxyPaX2voxy2+wE7HH43p4vI/zvQ1Um/11gyBhRzbMdz/xx4yfp8kYKwPq8EHsR9BXEn8Bn6d/2Lk76L+49+ALzosK+Nw+9QrOt8pcP+q4Fd7vz/KuPL+CAqnlD0MIiN36wygFfQT9DfishOEXkCQCm1HXgE/af4XUSmi0goF9MM/SRVtP+wcsochv4THLO2P0L/ybDePylS33GG1CkgEEBE6orIeyLym4gcRz+lNioyhnzQ4fNpJ9uBTuRrhn6y3lHSiVjXcgtws4jURT+BTrN2fwLMB6aLyH4ReVlEarroZ4tSKlUpFY5+mgxF3/SdIiKJIrLYcsjmAKMsud2hhTvnhn7aLgBCHOQcp7Qf4iv0DdTGniLyXWb9rvZZ382nTuRzbOP4uwXt28pz2LZ/70U4bL2HONnnjGeAB0XksmLqPA38Ff0bcIf/oq3ih7n4twvF/0dDufg62AhCW59rrMkBx4B5VnmVxCiIimc/cIXD9uVWGUqpXKXUo0qpVugb159tzmGl1DSlVHerrQJectL3IeC8k/73lVPmgcBapdRJa/tTYIA1G6Q9MMvNfh5FD5clKqUaAD2tcimnfIfQ4/5Xuln/M/RT+QBgs6U0UEqdV0r9XSnVAT1EdRMODk1XKKW2oq2JTrYiJ9Wmoa2MFkqphuihRCmmviN7gBKn81rfz0pgUEl1nRzz/6yySOu7uZOLv5cWDp/tv9tSsg19PoPdqWxd2/+iFYCrOgvQD1Z/dLPPU8BctNXhTEG4/I+ih/qKXgcbh9APMR2VUo2sV0OlJ31USYyC8C41RSTA4VUDfXN6WkSCRKQZ+gnpUwARuUlErhIRQTta84ECEWkrIteJSG30jdDmbC6EUiofbZJPEJH6InIF8Gdb/6VBNGEi8izaLH/K4Th70Q7cT4CZSqnTbnZb35L9mIg0AZ4trVzOUEoVoE3/10QkVET8ReRq63o5YzpwA/oGYbMeEJFeIhJpWTTH0cr2oussIu1E5FERCbe2W6AVzgqrykEgXERqOTSrDxxRSp0RkQTgdod92dZxXCmBb4AQEXlERGpb322ii7rjgJEi8oSIBFvyhQMtXdR3lO8EkCMiYWhHbFEeEpFw67v7K3oCQqlQeqzlz8DfRGSEiDQQET9r2vD7Lpr9HRgBNCqm67+iz91dngKuVUplOtnn8j+K/n+likgHywK1/4at3+G/gdcdrn2YiNxYCrkqFUZBeJc56Bui7TUePWMkAz3LZyOw1ioDaA0sRP9RfwTeVkotRjsOX0Q/oRxAO8+edHHM0ehx0J1oR9809M3TXUJFxDabZjXasZqklCoatPWRtc/ZE5grJqEdoofQN9N5pWhbEo+hr+dq9Eyfl3Dx+1ZKZaGvbzcK3+SaA1+ilcMW4Hucn18ukAisFJGT6HP5GW0hgZ4NtAk4ICKHrLI/As+JSC76hvO5gzyngAnAMmtoomsReXPRDuSb0d//r0AvF+f2A3r4pCfwi8MwRzp6Wq4r/o6eyZUDpKGf2osyDe3g3Yke8nrBSZ0SUUp9CQxD+3z2oxXqC2j/m7P6u9DfQ71i+lyGdji7K8N+61o5w+V/VCk1F/07/g5ttXxXpO3jVvkKa6huIdpqrpKI5TwxGEqFiPREP1VdocyPqNojIpnAvUqphb6WxVBxGAvCUGosx+2fgMlGORgM1RejIAylQkTao2czhVDMrB2DwVD1MUNMBoPBYHCKsSAMBoPB4JQKTdLlTZo1a6YiIiJ8LYbBYDBUKdasWXNIKeU0mK/aKIiIiAgyMjJ8LYbBYDBUKUTkN1f7zBCTwWAwGJxiFITBYDAYnGIUhMFgMBicUm18EAZDZeT8+fPs3buXM2fO+FoUwyVOQEAA4eHh1KzpNEGxU4yCMBi8yN69e6lfvz4RERHoHIwGQ8WjlOLw4cPs3buXli1Lytt4ATPEZDB4kTNnztC0aVOjHAw+RURo2rRpqS1ZryoIEekjIttEZLtt8RsX9QaLiBKReGv7DhFZ5/AqEJFob8pqMHgLoxwMlYGy/A69piCsnPpvodfs7QAMF5EOTurVRyd+W2krU0pNVUpFK6WigbvQS/at85asBh8zbRocPeprKQwGQxG8aUEkANuVUjuVUufQi7QMcFLveXTufle2z3CrraE6sns33HEHfPyxryWpdhw+fJjo6Giio6Np3rw5YWFh9u1z584V2zYjI4MxY8aUeIxu3bp5RNb09HRuuukmj/RVFsaPH2+/Pq1bt2bQoEFs3rzZZ/Lce++9Pj2+DW86qcMovHbrXvQiK3ZEJBa9BGOaiDhbwQr0wiLOFAsicj9wP8Dll1/urIqhsrNvX+F3g8do2rQp69Zpw3v8+PEEBgby2GOP2ffn5eVRo4bzW0B8fDzx8fElHmP58uWeEbYSMHbsWPv1mTFjBtdddx0bN24kKKjil5SePHlyhR/TGT5zUouIH/AaF1bhclYnETillPrZ2X6l1PtKqXilVLwvvkSDB8jKKvxu8CqpqamMGjWKxMRExo0bx6pVq7j66quJiYmhW7dubNu2DSj8RD9+/HhGjhxJUlISrVq14s0337T3FxgYaK+flJTEkCFDaNeuHXfccQe2TNFz5syhXbt2xMXFMWbMmFJZCp999hmRkZF06tSJxx9/HID8/HxSU1Pp1KkTkZGRvP766wC8+eabdOjQgc6dO3PbbbeV6zoNGzaMG264gWnTpvHdd99xyy232PctWLCAgQMH2s//r3/9K1FRUXTt2pWDBw8C8PXXX5OYmEhMTAzXX3+9vXz8+PHcc8899OjRgyuuuIL//ve/jBs3jsjISPr06cP58+cBSEpKsqcOmjdvHrGxsURFRdG7d+9ynVdp8aYFsY/Ci3uHW2U26qMXeU+3nCfNgdkikqKUsiVVug29PqyhumJTDPv3F1+vGvDIvHmsO3DAo31GN2/OpD59StVm7969LF++HH9/f44fP87SpUupUaMGCxcu5KmnnmLmzJkXtdm6dSuLFy8mNzeXtm3b8uCDD140n/6nn35i06ZNhIaGcs0117Bs2TLi4+N54IEHWLJkCS1btmT48OFuy7l//34ef/xx1qxZQ+PGjbnhhhuYNWsWLVq0YN++ffz8s35uPHbsGAAvvvgiu3btonbt2vay8hAbG8vWrVsZM2YMf/zjH8nOziYoKIgPP/yQkSNHAnDy5Em6du3KhAkTGDduHP/+9795+umn6d69OytWrEBEmDx5Mi+//DKvvvoqADt27GDx4sVs3ryZq6++mpkzZ/Lyyy8zcOBA0tLSCimj7Oxs7rvvPvv1O3LkSLnPqzR404JYDbQWkZbW4u23AbNtO5VSOUqpZkqpCKVUBHpdX7tysCyMWzH+h+qN7YZpLIgKY+jQofj7+wOQk5PD0KFD6dSpE2PHjmXTpk1O2/Tv35/atWvTrFkzgoOD7U/EjiQkJBAeHo6fnx/R0dFkZmaydetWWrVqZZ97XxoFsXr1apKSkggKCqJGjRrccccdLFmyhFatWrFz505Gjx7NvHnzaNCgAQCdO3fmjjvu4NNPP3U5dFYabBaQiHDXXXfx6aefcuzYMX788Uf69u0LQK1atewWUVxcHJmZmYBWwjfeeCORkZG88sorha5r3759qVmzJpGRkeTn59PHUvCRkZH29jZWrFhBz5497devSZMm5T6v0uA1C0IplSciDwPzAX/gA6XUJhF5DshQSs0uvgd6AnuUUju9JaOhEnAJWRClfdL3FvXq1bN//tvf/kavXr346quvyMzMJCkpyWmb2rVr2z/7+/uTl5dXpjqeoHHjxqxfv5758+fz7rvv8vnnn/PBBx+QlpbGkiVL+Prrr5kwYQIbN24spChGjBjBTz/9RGhoKHPmzCnxOD/99JPdDzNixAhuvvlmAgICGDp0qL3fmjVr2qePOp7z6NGj+fOf/0xKSgrp6emMHz/e3q/tOvn5+RVq7+fn57VrVla86oNQSs1RSrVRSl2plJpglT3jTDkopZIchpZQSqUrpbp6Uz5DJcCmIHJy4PRp38pyCZKTk0NYWBgAU6ZM8Xj/bdu2ZefOnfYn4xkzZrjdNiEhge+//55Dhw6Rn5/PZ599xrXXXsuhQ4coKChg8ODBvPDCC6xdu5aCggL27NlDr169eOmll8jJyeHEiROF+vvwww9Zt26dW8ph5syZfPvtt3aLJzQ0lNDQUF544QVGjBhRYnvH6/rRRx+5fc5F6dq1K0uWLGHXrl0AFT7EZFJtGHyL49BSVha0auU7WS5Bxo0bxz333MMLL7xA//79Pd5/nTp1ePvtt+nTpw/16tWjS5cuLusuWrSI8PBw+/YXX3zBiy++SK9evVBK0b9/fwYMGMD69esZMWIEBQUFAPzjH/8gPz+fO++8k5ycHJRSjBkzhkaNGpVK1tdff51PP/2UkydP0qlTJ7777rtCM5juuOMOsrOzad++fYl9jR8/nqFDh9K4cWOuu+46+w2+tAQFBfH+++8zaNAgCgoKCA4OZsGCBWXqqyxUmzWp4+PjlVkwqAoSEgI1asDevbB0KXTv7muJPMqWLVvcuqFUZ06cOEFgYCBKKR566CFat27N2LFjfS1WqXn44YeJiYnhD3/4g69FKTPOfo8iskYp5XROs8nFZPAd+fnw++8QG6u3jaO6WvLvf/+b6OhoOnbsSE5ODg888ICvRSo1cXFxbNiwgTvvvNPXolQoZojJ4Duys6GgAOLiYPbsS8JRfSkyduzYKmkxOLJmzRpfi+ATLnkL4kxeHj/s3k2Oyddf8dgsho4doWZNY0EYDJWMS15BZOzfT48PP+T731yu223wFjaFEBqqfRHGgjAYKhWXvIKIDQnBX4RVJhdQxWNTECEh+mUsCIOhUnHJK4i6NWsSedllRkH4AptCaN5cWxHGgjAYKhWXvIIASAwLY9W+fRRUkym/VYasLGjcGAICjAXhJXr16sX8+fMLlU2aNIkHH3zQZRvHRHH9+vVzmtdo/PjxTJw4sdhjz5o1q1DK6meeeYaFCxeWRnynmNTghfFmanCjIICEsDByzp7l18OHfS3KpUVWlrYeQFsQR4+aaGoPM3z4cKZPL5zObPr06W7nRJozZ06pA85sFFUQzz33HNdff32Z+qpsjB07lnXr1vHrr78ybNgwrrvuOrKzs30iy+TJk+nQ4aK12DyCURBoBQGw0gwzVSwHDmjLAS68ezjb6aXOkCFDSEtLsy8QlJmZyf79++nRowcPPvgg8fHxdOzYkWeffdZp+4iICA4dOgTAhAkTaNOmDd27d7enBQcd59ClSxeioqIYPHgwp06dYvny5cyePZu//OUvREdHs2PHDlJTU/nyyy8BHTUdExNDZGQkI0eO5OzZs/bjPfvss8TGxhIZGcnWrVvdPleTGtzzqcFNHATQvlkzAmvVYtW+fdwdFeVrcS4dsrLgmmv059DQC2VW5srqxiOPwDoPL5wbHQ2TJrne36RJExISEpg7dy4DBgxg+vTp3HrrrYgIEyZMoEmTJuTn59O7d282bNhA586dnfazZs0apk+fzrp168jLyyM2Npa4uDgABg0axH333QfA008/zX/+8x9Gjx5NSkoKN910E0OGDCnU15kzZ0hNTWXRokW0adOGu+++m3feeYdHHnkEgGbNmrF27VrefvttJk6c6NbiOSY1uHdSgxsLAvD38yM+NNRYEBWJUloZFLUgjKPa4zgOMzkOL33++efExsYSExPDpk2bih3HXrp0KQMHDqRu3bo0aNCAlJQU+76ff/6ZHj16EBkZydSpU12mDLexbds2WrZsSZs2bQC45557WLJkiX3/oEGDgMLps0vCpAb3TmpwY0FYJIaF8dqPP3ImL48AD/xgDCVw7BicPXtBMThaENWU4p70vcmAAQMYO3Ysa9eu5dSpU8TFxbFr1y4mTpzI6tWrady4MampqZwpY7Boamoqs2bNIioqiilTppCenl4ueW3psD2RMtykBi8fxoKwSAgL43xBAevNGHjF4BgDAdC0qU7aZywIjxMYGEivXr0YOXKk3Xo4fvw49erVo2HDhhw8eJC5c+cW20fPnj2ZNWsWp0+fJjc3l6+//tq+Lzc3l5CQEM6fP8/UqVPt5fXr1yc3N/eivtq2bUtmZibbt28H4JNPPuHaa68t1zma1ODeSQ1uHpUtHB3ViQ4phw1ewjEGAsDPz0x19SLDhw9n4MCB9qGmqKgoYmJiaNeuHS1atOAamy/IBbGxsQwbNoyoqCiCg4MLpe1+/vnnSUxMJCgoiMTERLtSuO2227jvvvt488037c5pgICAAD788EOGDh1KXl4eXbp0YdSoUaU6H5Ma/GK8kRrcpPt2IOy11+gVEcGn1hiowYtMnQp33glbtkC7drosMREaNoRvv/WtbB7EpPu+NKmsqcErVbpvEekjIttEZLuIPFFMvcEiokQk3qGss4j8KCKbRGSjiAR4U1bQVoRxVFcQRYeYbJ+NBWGo4lSn1OBeG2ISEX/gLSAZ2AusFpHZSqnNRerVB/4ErHQoqwF8CtyllFovIk2B896S1UZiWBiztm7lyOnTNKlTx9uHu7TJyoI6dcCaVQJoR/XSpb6TyWDwANUpNbg3LYgEYLtSaqdS6hwwHRjgpN7zwEuA4xSKG4ANSqn1AEqpw0qpfC/KqgW2/BCrjRXhfWxTXK1ZGYDePnJEz24yGAw+x5sKIgzY47C91yqzIyKxQAulVFqRtm0AJSLzRWStiIxzdgARuV9EMkQkwxNh7vGhoQgmorpCcIyBsHEJTHU1GKoSPpvmKiJ+wGvAo0521wC6A3dY7wNF5KK4caXU+0qpeKVUvOMMgrLSoHZt2gcFmcyuFYFjHiYbNoVhFITBUCnwpoLYB7Rw2A63ymzUBzoB6SKSCXQFZluO6r3AEqXUIaXUKWAOEOtFWe3YHNXVZXZXpcUxD5MNmwVhYiEMhkqBNxXEaqC1iLQUkVrAbcBs206lVI5SqplSKkIpFQGsAFKUUhnAfCBSROpaDutrgQrJp5sYFsahU6fI9EB+FoMLTp+GnJyLFYSxIDzK4cOHiY6OJjo6mubNm9tTVEdHR9uT97kiIyODMWPGlHiMbt26eURWX6fw3rZtG0lJSURHR9O+fXvuv/9+ALeD5YrWmz17Ni+++KLX5K0ovDaLSSmVJyIPo2/2/sAHSqlNIvIckKGUml1M26Mi8hpayShgjhM/hVewOapX7dtHy8aNK+KQlx7OprgCBAWBv79REB6iadOmrLOyA44fP57AwEAee+wx+/68vDyXeYji4+PtqSOKY/ny5Z4R1seMGTOGsWPHMmCAnkezceNGQN/4MzIy6NevX7Hti9ZLSUkplK+qquJVH4RSao5Sqo1S6kql1ASr7BlnykEplWRZD7btT5VSHZVSnZRSTp3U3iAyOJiAGjXcdlSfOAEPPGDuaaXClYLw89N+CTPE5DVSU1MZNWoUiYmJjBs3jlWrVnH11VcTExNDt27d7Gm8HZ/ox48fz8iRI0lKSqJVq1a8+eab9v4CAwPt9ZOSkhgyZAjt2rXjjjvusA/Tzpkzh3bt2hEXF8eYMWNKZSlUVArvrKysQpHZkZGRnDt3jmeeeYYZM2YQHR3NjBkznF4vZ/WmTJnCww8/DMDBgwcZOHAgUVFRREVFVSmlalJtFKGmvz+xISFuO6r/9z94/3248koYV2FqrIrjSkGA9kNUV2275hE46uF8342jIa50WQD37t3L8uXL8ff35/jx4yxdupQaNWqwcOFCnnrqKWbOnHlRm61bt7J48WJyc3Np27YtDz74IDVr1ixU56effmLTpk2EhoZyzTXXsGzZMuLj43nggQfsKajdXagIKjaF99ixY7nuuuvo1q0bN9xwAyNGjKBRo0Y899xzZGRk8K9//QvA5fUqWm/KlCn2vseMGcO1117LV199RX5+/kV5nyozJlmfExJCQ1mTlcX5/JJDL2zDjuVMeXJpUTQPkyMhIcaC8DJDhw7F398f0Inihg4dSqdOnRg7dqzLVN39+/endu3aNGvWjODgYPuiNo4kJCQQHh6On58f0dHRZGZmsnXrVlq1amVPQV0aBVGRKbxHjBjBli1bGDp0KOnp6XTt2tW+iJEj7l4vR7777jv7Eq/+/v40bNiwVLL5EmNBOCExPJxJK1fy8++/E+PsKdciPx/mzdOxXkuXat+rCcB2g6ws7WtwNjU5NBSqkAleKkr5pO8t6tWrZ//8t7/9jV69evHVV1+RmZlJUlKS0za2tNPgOg23O3U8gbdSeIeGhjJy5EhGjhxJp06d7BaKI+5er+qCsSCc4O4SpCtX6sDf1FQd/GuyRLjJgQNw2WXa51CUkBA4dAhKmGVj8AyOqaYdh0U8Rdu2bdm5c6d9cZsZM2a43bYiU3jPmzfPvoTngQMHOHz4MGFhYRelLHd1vVylNgfo3bs377zzDqD9Jzk5OW5fA19jFIQTWjZqRLO6dUv0Q6Sl6QfhF16AWrXMMJPbOIuitmHWpq5Qxo0bx5NPPklMTIxXnvjr1KnD22+/TZ8+fYiLi6N+/fouh1hsKbxtr8zMTHsK76ioKOLi4hgwYAD79u2zT0m98847C6XwjoyMJCYmptQpvL/99ls6depEVFQUN954I6+88grNmzenV69ebN682e58dnW9itZz5I033mDx4sVERkYSFxdX7Mp9lQ2T7tsF/adN47djx/j5j390WScmRuea+/576NULjh71/JrD1ZLoaGjRAhwWnbGTlgY33QQ//ghdu1a8bB7GpPuGEydOEBgYiFKKhx56iNatWzN27Fhfi3VJUqnSfVcJjq6DeQlwZG2h4sSwMDZnZ3PcReK4ffu0MujfX28nJ8P69eDEd2coijsWRHWdyXQJ8u9//5vo6Gg6duxITk4ODzzwgK9FMriJURC1m8GR1fD794WKE8LCUMAaFzNqbCs02uJnbrhBvy9a5CU5qwt5eZCd7VpBmHQb1Y6xY8eybt06Nm/ezNSpU6lbt66vRTK4iVEQdcOhXkv4vbCHuYt1o3LlqE5Lg8svh44d9XZMDDRpYvwQJXLwICjlfIor6JlNfn7GgjAYKgFGQQAE94TspfrGZdG0btOoppYAACAASURBVF2uatLEqaP67FmtCPr1u7Ccgb8/9O6ty6uJW8c72JzPriwIf3+tPIyCMBh8jlEQAME94OwhOL61UHFCWJhTBbF0KZw8ecH/YCM5Wfsmtm69qInBRnFR1DZMsJzBUCkwCgIgqKd+/31JoeLEsDD25eay7/jxQuVpaVC7Nlx3XeFukpP1uxlmKgZ3FER1TrdhMFQhjIIAqH8VBDS/SEE4ZnZ1ZM4cPa21qK8tIgKuusooiGIpLs2GDWNBeIxevXoxf/78QmWTJk2yp35wRlJSErYp4/369XOa12j8+PFMnDix2GPPmjWr0Jz/Z555hoULF5ZGfKeY1OAVh1EQoB0JwT0ge0khB0J08+bU9PMr5Kjevh1++eXi4SUbycmQng5WUKahKFlZ0LSpjix0RWionulkLmK5GT58ONOnTy9UNn36dLdzIs2ZM6dUAWeOFFUQzz33HNdff32Z+qpM2FKDr1u3ji1btjB69Gig7AoiJSWFJ554wmvylgejIGwE9YRTe+Hkb/aigBo1iGrevJAFYfteXaWHT07WKcBXrPCmsFUYZ0uNFsVEU3uMIUOGkJaWZl8gKDMzk/3799OjRw8efPBB4uPj6dixI88++6zT9hERERw6dAiACRMm0KZNG7p3725PCw46zqFLly5ERUUxePBgTp06xfLly5k9ezZ/+ctfiI6OZseOHaSmpvLll18COmo6JiaGyMhIRo4caU+MFxERwbPPPktsbCyRkZFsLYVDz6QG9zwmWZ+N4B76PXspBEbYixNCQ/lkwwbyCwrw9/MjLQ3atYNWrZx306uXnqW5YAH06OF9sasczpYaLYpjsFyLFsXXrUo88ojnQ+2jo2GS6ySATZo0ISEhgblz5zJgwACmT5/OrbfeiogwYcIEmjRpQn5+Pr1792bDhg107tzZaT9r1qxh+vTprFu3jry8PGJjY4mLiwNg0KBB3HfffQA8/fTT/Oc//2H06NGkpKRw0003MWTIkEJ9nTlzhtTUVBYtWkSbNm24++67eeedd3jkkUcAaNasGWvXruXtt99m4sSJTJ48ucTLYFKDewdjQdho2AlqNrrYUR0eTu65c2w9dIiTJ/XwUXGLSzVqBAkJxg/hkuKiqG2YYDmP4jjM5Di89PnnnxMbG0tMTAybNm0qNkfQ0qVLGThwIHXr1qVBgwaFVkv7+eef6dGjB5GRkUydOrXEFNjbtm2jZcuWtGnTBoB77rmHJUsu/O8GDRoEQFxcnD3JX0mY1ODewasWhIj0Ad5ALzk6WSnl1BMjIoOBL4EuSqkMEYkAtgA2O3aFUmqUN2XFzx+CrinWUb1jTzDnzrn2P9hIToYJE+DYMa0wDBZKld6CqE4U86TvTQYMGMDYsWNZu3Ytp06dIi4ujl27djFx4kRWr15N48aNSU1N5cyZM2XqPzU1lVmzZhEVFcWUKVNIT08vl7y2tOGeSBluUoOXD69ZECLiD7wF9AU6AMNFpIOTevWBPwEri+zaoZSKtl7eVQ42gntC7i9w+kJCpTZNm9Kwdm1W7ttHWhrUrw/duxffTXIyFBTAd995Wd6qxpEjOo13SQoiOFiP0xkLwiMEBgbSq1cvRo4cabcejh8/Tr169WjYsCEHDx5kri13jAt69uzJrFmzOH36NLm5uXztkGgxNzeXkJAQzp8/z9SpU+3lrlJgt23blszMTLZv3w7AJ598wrXXXluuczSpwb2DN4eYEoDtSqmdSqlzwHRggJN6zwMvAWV7fPEkwVY8RPaFtBt+InQJC2Pl3n3MmaNv/sVNwAGdhDQw0AwzXYQ7MRAANWpoJVHdLAgfMnz4cNavX29XEFFRUcTExNCuXTtuv/12rrnmmmLbx8bGMmzYMKKioujbty9dunSx73v++edJTEzkmmuuoV27dvby2267jVdeeYWYmBh27NhhLw8ICODDDz9k6NChREZG4ufnx6hRpXsGNKnBKwillFdewBD0sJJt+y7gX0XqxAIzrc/pQLz1OQI4CfwEfA/0KOl4cXFxqtzknVVqeh2lVo8pVPzUwoXK78F3FCg1ebJ7Xd10k1JXXll+kaoV336rFCiVnl5y3dhYpfr29b5MXmbz5s2+FsFgsOPs9whkKBf3VZ85qUXED3gNeNTJ7izgcqVUDPBnYJqINHDSx/0ikiEiGdnZ2eUXyr8WNLtax0M4kBgeTsEvVwHQt697XSUnw44dsGtX+cWqNpSUh8mRkBBjQRgMPsabCmIf4DhHMdwqs1Ef6ASki0gm0BWYLSLxSqmzSqnDAEqpNcAOoE3RAyil3ldKxSul4oOcrW9cFoJ6wNH1cO7C2F9CWBj82pqwNifsE2xKwqTdcIK7Q0xg0m0YDJUAbyqI1UBrEWkpIrWA24DZtp1KqRylVDOlVIRSKgJYAaQoPYspyHJyIyKtgNbATi/KeoHgnoCC7GX2otrnA2FPCxpG7nG7m3btIDzcKIhCZGVBvXra018SISHw++96/YgqjjLpfQ2VgLL8Dr2mIJRSecDDwHz0lNXPlVKbROQ5EUkpvjU9gQ0isg49/XWUUuqIt2QtRLOuIDUKDTPNnw8oP4622OB2NyLaili0CPLzvSBnVcSdGAgboaF6WmwVX6IvICCAw4cPGyVh8ClKKQ4fPkxAQECp2nk1DkIpNQeYU6TsGRd1kxw+zwRmelM2l9SoC03iCy0gNGcO1Gt0nqwG28g+eZKgevXc6io5GT78ENauBYdJH5cupVEQtnr794M1VbAqEh4ezt69e/GIj8xgKAcBAQGFUoS4g0m14YzgnrDtdcg7Tb7UYe5c6NbrLAv8FKv27aN/m4vcIU7p3Vu/L1hgFASgFURMjHt1bc6eKu6HqFmzJi1btvS1GAZDmTCpNpwR3AMKzsPhlaxeDYcOwfCBAfiJOF1AyGU3wTpVjvFDWLiTqM+GowVhMBh8glEQzgi6BhD4fQlz5uig3gH9a9ApONjlGtWuSE6GZcv0CnSXNCdO6Je7Q0yXXaYdOVXcgjAYqjJGQTijVmNoFAnZS0lLg6uvhiZNdGbXVfv2lcrhmJyslzVYsqTkutWa0sRAwIVoamNBGAw+wygIVwT3pOD35WxYf96enC8xPJyjZ86w/Yj7E6q6d9fLk17yw0yliYGwYYLlDAafYhSEK4J74ldwitiItfb03q6WIC2OOnX0uhBGQZRBQYSGGgvCYPAhRkG4Ikiv9nNTwlJsa6h0DAqiXs2apVIQoIeZfv75Er/XGQvCYKhyGAXhgnP+zdl+sDUpXZcgosv8/fyICw0tk6MawAPrtVddsrK0X6FJE/fbhIZWm2hqg6EqYhSEC5Ytg++39KB9sx9AFdjLE0JD+enAAc6VIjw6KgqCgi7xYaYDB/QUV79S/ORCQvTCGr//7j25DAaDS4yCcEFaGizf3pNaHIWcC0sFJoaHcy4/n/W2WTlu4Oeng+YWLtTZIy5JShNFbaOaBMsZDFUVoyBckJYG5xtpP4Rj2o2yOKpBDzMdOKB9EZckZVEQJljOYPApRkE4YedO2LoV4q5tCXXCCq1T3aJBA5oHBrKqlDetSz79t7EgDIYqh1EQTrAtQ9uvn+i8TNlL7GNDIkJCWBgr9+4tVZ8tWugU4Jekgjh3TucrKa2CsEVTGwvCYPAJRkE4Yc4caN1avwjuAaez4MSF5SgSQkPZdvgwx86Ubhnt5GT4/ns4e9bDAld2bCm73c3DZKNmTWjWzFgQBoOPMAqiCKdOweLF2IPjCOqp3x2GmRKtlLmry+CHOH0ali/3hKRViNKm2XDEBMsZDD7DKIgiLF4MZ85gT69Bw/ZQqwlkX3BUx1tj46V1VCcl6VCAS26YqSxBcjZMsJzB4DOMglAKdu2yb6al6VUxe1qGA+Knh5kcLIhGAQG0a9as1I7q+vWha1ejIEqFsSAMBp9hFMS6ddCqFXTsiHr8CQ7+dxk39M6ndm2HOkE94cQOOHXhRmVzVJd2KcnkZFizBg4f9pD8VYGsLO1svuyy0rcNCdE+DLNuq8FQ4XhVQYhIHxHZJiLbReSJYuoNFhElIvFFyi8XkRMi8pjXhAwLg9df1w7U115l5sHuTFt0Gdx9N3zxBRw/ri0IKDTMlBgWxsGTJ9lz/HipDpecrI2W777z5ElUcrKytLO5Zs3Stw0NNdHUBoOP8JqCEBF/4C2gL9ABGC4iHZzUqw/8CVjppJvXgLnekhHQaw488ggsWsSbT2czjOnk39hXjzXdequ+sQ17Ar6tBT99bW9mC5gr7XTXLl2gYcNLbJipNCvJFcU2LGX8EAZDheNNCyIB2K6U2qmUOgdMBwY4qfc88BJQaM6oiNwC7AI2OWnjFb5a3IhtUcOoN/MTPayxZIlWHvv2w0fnYOBU6NgRHn+cqF9/pU4plyAF7aTu1Qu+/fYSSrtRliA5GyZYzmDwGd5UEGHAHoftvVaZHRGJBVoopdKKlAcCjwN/L+4AInK/iGSISEZ2dna5hD12DH74wWF6a40aeiGHl1+GLVsg7RG4EwhuCq+9Rs2kJLJefpnrx4+Hr74q1d0+ORl++w22by+XyFWHAwfKriBMug2DwWf4zEktIn7oIaRHneweD7yulDpRXB9KqfeVUvFKqfigoKByybNggfaD2qe3FiXuFj1Y9slfdFTwjBn8kpBA/Lp1MGiQVhJuckml3SgoKJ+CsA1NGQvCYKhwvKkg9gEtHLbDrTIb9YFOQLqIZAJdgdmWozoReNkqfwR4SkQe9qKspKVB48aQmOiiQtME8Kul0240bAi33sqvb7xB88ceI69hQ92Bm1x1FVxxxSWiIA4f1us5lFVB1Kql/UDGgjAYKpwaXux7NdBaRFqiFcNtwO22nUqpHKCZbVtE0oHHlFIZQA+H8vHACaXUv7wlaEEBzJ0LffrokSWn1KijlYRDPERCWBh5/v7siY2l5YIFepjJtrpQMYhoK+Lzz/W90+UxqwPliYGwYYLlDAaf4DULQimVBzwMzAe2AJ8rpTaJyHMikuKt45aFNWv0LEq7/8EVQT3gyFo4r0e+rmzcmCZ16rCsbVvYswd+/dXtYyYn6xm0q1eXQ/CqgCcUhAmWMxh8gld9EEqpOUqpNkqpK5VSE6yyZ5RSs53UTbKsh6Ll45VSE70p55w5+qm+T58SKgb3BJUHh1cAFzK7TrPd/EoxZtS7tz5mtR9msimIsk5zBWNBGAw+wkRSo90HiYl6qLtYgrrp1BuOCwiFhjIfKIiIKNXdvmlTiIu7BBREeRL12QgN1f2YaGqDoUK55BXEwYN6mMfl7CVHajaARtEXZXYtUIp9iYk6019entvHTk6GFSsgN7cMglcVsrJ0Eqp69creR0iIVg6HDnlOLoPBUCKXvILYtw86d3bD/2AjuIceYso/B8C1V1xBeIMGTKpfv9ROheRkrU/S00svd5WhPEFyNmzBcsYPYTBUKJe8goiNhfXr9btbBPeE/DNwRLtL6tWqxbv9+zOlSRNUKZ0K3bpB3brVfJjJEwrCpNswGHzCJa8gSk2QNQPXYZipf5s23JiQwNqQEE7Z1it1g9q1dVpxoyBKwKTbMBh8glsKQkTqWZHPiEgbEUkRkTKk5qwGBARBg3aFMrsCvNGnD0vbtKFWRgYFpcjwmpwMW7dCKXP+VQ2UKl+iPhu29maIyWCoUNy1IJYAASISBnwL3AVM8ZZQlZ7gnpD9AxRcmFUTVK8eHW+/nRr5+cyZPNntrm64Qb/P9W7OWt9w4oRew7W8FkTt2nral7EgDIYKxV0FIUqpU8Ag4G2l1FCgo/fEquQE9YDzxyFnY6Hi6++8kzO1arH7yy/ZnZPjVlcdO+q0G19/XXLdKocnguRshIQYC8JgqGDcVhAicjVwB2BLOuTvHZGqAMHWeqQOfggAqVMH1b07123fzqhvvnFrtTkRSEnRfohTp7whrA/xpIIIDTUWhMFQwbirIB4BngS+stJltAIWe0+sSk69y6HeFYUC5mzU6duXdtnZrF+7lmkbNzppfDEpKXDmDCxc6GlBfYyxIAyGKo1bCkIp9b1SKkUp9ZLlrD6klBrjZdkqN0E9dGbXolaClct71NGj/GnePLJPniyxq549oUEDmH1RApIqjqcVxIEDOrOiwWCoENydxTRNRBqISD3gZ2CziPzFu6JVcoJ7wpnfIfeXwuWRkRAczOjjxzl+9ixj588vsatataBvX+2HqFb3v6wsfXKNG5e/r9BQHVVooqkNhgrD3SGmDkqp48At6DWiW6JnMl262OMhigwz+flB7940+uEHnurenakbNzLXjSyvKSk6o+yqVV6Q1VfYpri6kQK9REywnMFQ4birIGpacQ+3ALOVUueBS2VFZec0aAu1gy5yVAN6mOngQZ5q1IgOQUE88M035J49W2x3ffuCv381G2Yqz0pyRTHpNgyGCsddBfEekAnUA5aIyBWA+9Fg1RERnZcp+2JHNddfD0CtxYuZfPPN7D1+nKcWLSq2u8aNtS+iWikIT0RR2zAWhMFQ4bjrpH5TKRWmlOqnNL8BvbwsW+UnqCeczISTuwuXt2gBbdvCggVc3aIFoxMSeGv1apbv2VNsdykpsGkT7NjhPZErFG8oCGNBGAwVhrtO6oYi8pqIZFivV9HWxKWNPR7CiRWRnAzffw9nzzKhd29aNGzIvbNnc7aYdOA336zfq0XQ3NmzcOSI5xREQIA2s4wFYTBUGO4OMX0A5AK3Wq/jwIclNRKRPiKyTUS2i8gTxdQbLCJKROKt7QQRWWe91ovIQDflrFgaddZrRDgbZkpOhtOn4ccfCaxVi/duuokthw4xYamTuhZXXqkjq6vFMJNtoaDy5mFyxATLGQwVirsK4kql1LNKqZ3W6+9Aq+IaiIg/8BbQF+gADBeRDk7q1Qf+BKx0KP4ZiFdKRQN9gPdEpIabslYcfv7Q7BrnjuqkJO11tlK19rnqKu7s3Jl//PADGw8edNllSgosWQJHj3pJ5orCkzEQNkywnMFQobirIE6LSHfbhohcA5wuoU0CsN1SKOeA6cAAJ/WeB14CztgKlFKnlFK2sZgAKvOMqeAecHwLnMkuXN6ggV7H1CGX9+s33kijgADu/fpr8l0EPKSk6MXTqnzyPk8sNVoUY0EYDBWKuwpiFPCWiGSKSCbwL+CBEtqEAY5e2b1WmR0RiQVaKKXSKIKIJIrIJmAjMMpBYTjWud/mF8nOzi66u2IITtLv29+/eF9yMmRk6LF4oFndurzZpw+r9u3jny4CHhISIDi4GgwzecuCyMq6OHrdYDB4BXdnMa1XSkUBnYHOSqkY4LryHNhK2fEa8KiLY65USnUEugBPikiAkzrvK6XilVLxQUFB5RGn7DTrCpffChufgaxvC+9LTtY3s8UX0lbd1qkT/Vq35q/ffccuJ+NIfn7aWT13Lpw7523hvUhWljUVONhzfYaEwPnzcPiw5/o0GAwuKdWKckqp41ZENcCfS6i+D2jhsB1uldmoD3QC0i2rpCsw2+aodjjmFuCEVbfyIQJdP4CGHWHZbZDrMEc1IQHq1y80zCQivNO/P34iPOAi42tKil7eeokT10aVIStLK4caHnQdmWA5g6FCKc+SoyXlT1gNtBaRliJSC7gNsA+cKKVylFLNlFIRSqkIYAWQopTKsNrUALCC8tqhA/UqJzXqQc9Z+vPSgZBnJeirWVM7q4usKXp5w4a82Ls3C3bu5OP16y/q7vrr9azOKj3M5MkYCBuVIViuoEAff8UKmDEDXnkF/vxn+OWXktsaDFWM8jzeFTsQrJTKE5GHgfnotSM+sFKFPwdkKKWKu/11B54QkfNAAfBHpVTlztIW2AqumQ7pfWHFSP1ZRA8zff017NwJrS5M/HqwSxc++/lnxs6fT5+rruKywED7vrp1dbPZs+GNNzyTyqjC8cRSo0WpCAvi3DnYswd++w1279bvttfu3frlbOzv4EGYOtV7chkMPqBYBSEiuThXBALUKalzpdQcYE6Rsmdc1E1y+PwJ8ElJ/Vc6Qm6AqP+DdU9Ak3jo8Bd7+m8WLoT777dX9RPh3zffTPR77zFm3jxmDBlSqKuUFK1XNm6Ezp0r8iQ8xIEDEBXl2T69YUEUFMCf/gRr1mgl4MwJHhKil/2LjYWBA/XnK66Ayy/X708+CVOmQE4ONGzoOdkMBh9TrIJQStWvKEGqDe3HwZE1sP4JaBwNba+HsDA9zOSgIADaBwXxdI8ePJOeTvtmzfhbz574++lRv5tu0nVmz66CCiI/Xz9Re3qIqU4daNTIsxbEjz/Cv/4FXbroBcJtN3+bAmjRQq+JXRypqfDOO/DFF3DvvZ6TzWDwMZUv+KyqIwKJH8DxrbBsGPTJ0FbE//6nb5z+hVdqfbx7d345coS/f/896ZmZTB00iLAGDWjeXIdRzJ4NTz/to3MpK4cO6XP1tIKAC1NdPcXMmXrNioULdexKWejSBdq311aEURCGakR5nNQGV9QMhB5f6c9LBsJ1PXVo9E8/XVS1lr8/H99yC1MGDGD1/v1Evfsu31gOz5QUWL26Ck7a8UYMhA1PBssppRXEDTeUXTmAfii45x5YtgzcWPvDYKgqGAXhLepfCd0+g2MboeH/dFmR2Uw2RIR7oqNZe//9hDdowM2ffcbYefPo00/HBn7zTUUJ7SG8qSA8mW4jI0M7nYv4f8rEnXfqIJaPPip/XwZDJcEoCG8SeqN2Wp/4H7QJcakgbLRt1owV997L6IQEJq1cyb0rPyD8ivyqN93VpiA8PYsJLlgQnoimnjlTx2mkpJS/r7AwbYl8/HE1WzfWcCljFIS36fA4tBgCrbNg2Q9w6lSx1QNq1ODNvn2ZNWwYmceOcjBsLd8uKODkyQqS1xN4Iw+TjZAQPc3USl9SZpSCL7+E667zzJrZoJ3Ve/YUipw3GKoyRkF4GxHo+iEkXAHnzsO3X7jVbEC7dqwfNYq21xzi/Dk/bn5hJSeqSu6NrCw93bNOiTOhS48tFqK8fogNG/TKTJ4YXrIxYIA+7ylTPNenweBDjIKoCGoGwgPf6DljHz92IdK6BFo0bMiq528kIPA8i+fVIu7991lnezr3FAcPwtVXw8qVJdd1F29EUdvw1MpyM2dqn8Ett5RfJhsBATB8uO77+KW9Iq+hemAUREVxWSdIiII1h2DlvW6Podep7ceglJo02tOZ3DPnSZw8mX+tWuU0h1OZeOEFnTbin//0TH9QMQqivBbEl1/qRcA9neTxnnv0QlFfuGcpGgyVGaMgKpKbhsFuYMN02Pqq281SUuDYYX8mxz5IcqtWjJ47l4EzZnC4BH9GiezcCe+9p598Z83CY46Oym5BbNmiX4MHe0YmRxIT9XrkZpjJUA0wCqIisaXdyEqAdY/DgYVuNevTR0+2WfJtAF8PH87rN97InF9/Jfq991jy229ll+eZZ3Tg3gcfaOXgicWwlfJOHiYb9erpmIXyWBAzZ+r3QYM8I5MjItpZ/cMPsH275/s3GCoQoyAqkpgYPWPmt9bQoD38MAxO7CqxWcOGOins7Nk6ZuKRrl358Q9/IKBGDXp99BHPff89eaWdWrl+PUybxrkH/8RfNwwjr3kYTJtWtvNyJCcHzpzxngUB2lFdHgviyy+hW7cLDm9Pc9dd2r/x8cfe6d9gqCCMgqhI/P2hd29YlK4jrVWBjrTOK3moKCVFj4rYAnXjQkNZe//9DO/UiWfT00mcPJnV+/YV34kjf/0rqmFDHtr9OP/3oh/zGw/XqxSVdzEeb05xtVGedBs7dmjl6I3hJRthYdpa/OgjExNhqNIYBVHRJCfDvn2wPx+umQbHNrjltL75Zv3uOApUv3ZtPhk4kBlDhpCVm0vi5MmMnjOHnDNnnHdiY+lSSEtjRdITTJ7ZmFat4Nltt0Nenn66Lg/ejKK2UR4Lwja85E0FAXqYafduSE/37nEMBi9iFERFY/NDLFgAoX0h6gX47TNYcguc3OOyWUSEzupaNKpaRLi1Y0e2PPQQD3XpwlurV9P+rbf4fNMm5zOdlIInnuBcUCh900bTt6++h23wiyarcfvyDzNVhIIoz9rUM2dCfLzO1upNTEyEoRpgFERF07KlXjjIlnajw5MQMxEOLIC0DrDtn1CQ77RpSor2fTobBWoYEMA/+/Vj1X33EVK/PsO+/JJ+06axo2jE8TffwPLlPJv/LA1D6vLJJzqj9d33CO/l3q7XOd3jWlGVSEVZEGfPwrFjpWu3ezesWuV96wF0kOCwYVoh5eZ6/3gGgxcwCsIXJCfrx/bz5/Wsl/aPQv9NEHQNrBkDC7rB0Q0XNUtJ0Vm058513XV8aCir7r2XN/r0Ydnu3XR65x0mLFnCufx8yM9HPfkk++q25o3jI/jyS2jaVLcbNw4+zRuuN6ZPL/u5ZWXpabPeXDinrFNd//tf/V4RCgL0MNOpU+UftjMYfIRXFYSI9BGRbSKyXUSeKKbeYBFRIhJvbSeLyBoR2Wi9X+dNOSuc5GT9VLlq1YWywJaQNBe6TdMzm+bFwbonIe+0vUpcnL43lpS8z9/PjzGJiWx56CFuatOGpxcvJvrdd9k6aRKyaRNjT73AxDdq0qXLhTZt2kDMkCtZ7Z9I/iflGGayTXH15jqpZU23MXOmHqdr3drzMjmja1d9Yc0wk6GK4jUFISL+wFtAX6ADMFxEOjipVx/4E+CY6+EQcLNSKhK4h6q4/Ghx9Oqlb6BFs7uKQMRwuGkrtLwbNr8IcyLt8RJ+ftpZPW+eHmEpibAGDfhi6FDSbr+d/DNnCHj2H2QQR8GQW3jwwYvrP/EEfJJ/O/4b18HmzWU7twMHvDu8BGWzILKy9HoNFWU9wIWYiCVL9Owpg6GK4U0LIgHYrpTaqZQ6B0wHBjip9zzwEmCfeqOU+kkpZfv3bwLqiEgJ6z5WIZo00Y5SV+m/azeBrv+B3t+B+MF3ybD8bjiTTUqKNj6+/979w/Vr3Zplx4WIk4d5wbBA1QAAIABJREFUsuHjfNf5LT5c9xMFRZy8cXFwoOcw8vEj7+MyWhHejKK2UZZ0G199pZ3ankzO5w533aUVhYmJMFRBvKkgwgBHb+deq8yOiMQCLZRSacX0MxhYq5S66JlZRO4XkQwRycjOzvaEzBVHcrJOkJeT47rOZb2g3wbo9DfYPR3S2pN81cfUratKtUbE+SO51Pi/f7DYrzejv7mOjuGN+MPs2Vw7ZQqbfv+9UN1Rz17GQq7n1H+mlW2WUEUoiMBAqF+/dBbEzJnQrh10uMiI9S7h4XD99WadCEOVxGdOahHxA14DHi2mTke0dfGAs/1KqfeVUvFKqfggTydd8zbJydrjXJIp4B8AnZ+DPj9B/bbUWnMPy19IZv2y7W7fvxff/BqN8g5xdvw/SOnelO9TU/lPSgqbs7OJfu89nly4kN+tPEy9esGKlrfT4NAu8paVMsPr6dN6ZpG3FQSULlguO1tPCqjI4SVHUlMhM1MPNRkMVQhvKoh9QAuH7XCrzEZ9oBOQLiKZQFdgtoOjOhz4CrhbKVX9BnCvvhrq1i1xlTk7jTpC8lLo8jbtg1fz7Z8iyVr0IhScL7ZZ2oe/c/XyifzUajB9/qa90n4ijIyJYdvDD3Nn5868uGwZIa++yrVTpvDPVSu54skbOENtdk4o5TBTRURR2yhNsNz//qef3it6eMnGLbfo/FHGWW2oYnhTQawGWotISxGpBdwG2AdGlFI5SqlmSqkIpVQEsAJIUUpliEgjIA14Qim1zIsy+o7atXW6aXcVBGh/ROsHOd5zC2nr+hP6+5N6ttMh50/6O3bAb6P+jzqcpuOsCRftb1a3Lh8OGMCGUaN4ukcPDp86xZ/mzWPE3veZU7sPjRdM59eDvzvp2QXeXGq0KKWxIGbO1LEnUVHelckVdevqmIgvv4QTJ3wjg8FQBrymIJRSecDDwHxgC/C5UmqTiDwnIiUtAvwwcBXwjIiss17B3pLVZyQnw7ZtpQ5Ma9YilNdXf8mj/5sFZ4/At1fDdzfArk/gvL4BnT4No2/O5A/n3uH0bSOoFdnWZX+Rl13G33v14uc//pGtDz3EP5J7M79bHEH52Tw8fCKd33mHv6ens/HgweLXofCFBVHSONvRo7BokR5e8ubU25JITdUZc01MhKEKIR5beMbHxMfHq4yMDF+LUTo2btTz8j/4AEaMKFXTl17S01L37solLPc12PkRnNwFNepB+CBe++oumk38hDtqfYH/jl+1s7QUnMs9y+mGl7GoUT8mvZrMD7t3o4DWTZowqH17BrdvT3xoKOJ4033rLXj4Ye+m+7bx6qvw2GNaATRq5Lrexx/rRXxWrNBrNfgKpfQ6EWFhZs1qQ6VCRNYopeKd7TOR1L6kUye47LLSDTNZpFg22Nfz6kPks5CyA65fChF3cHbX1/y52Q3cJZ/gf1sk1DtSfGdOqFW/NnsShpB89GtejriNrEcf5d3+/WnZuDGv/vgjCZMnc8WkSTwybx5LfvtNT5nNytLBGhUxYcDdYLmZM3UukYQE78tUHCJaUaWnw66SU7wbDJUBoyB8iYieArlwYamnQLZrB1dd5RBVLQLB3Vlf+z0uG5XF1rfjoW4N6L4W5kbBnGjY8iqcdj924Mq/3U59TvD9uG+4LDCQB+LjmX/nnRx87DE+uuUWYkJCeDcjg2unTOGKSZNYlZHB+aAgnda8FBw4UIYZte4Ey+Xmwvz5emEgXw4v2TAxEYYqhlEQviY5WU/DnDevVM1EtBWxaNGFXHA5OXqizrW119Budwby1N/hrgMQ/y/wrw0/PQazwuG7G2HXp5BX/BKjdfpcS25gCG0ypvHzzxfKm9Spw91RUfzvtts4NG4c0wcPJqZ5cw7v3MkGPz86v/MOL/3w/+2deXhU1fnHPyd7SCBAkABBVlkECrIEWzdwAZdWq6AsKoilWi1Yte4/rcatVq27rXWjolIVBFcU2UVBRUBAdtlkTySQfZ95f3+cG2YymQkhycwQeD/Pc5577rnn3vvOyeV8Odt7vmZndWs8HKZNs42BG244QpGoSQti5ky75Dxc01t9adfO7gei+0QoDQUROSZC//79pUHyyy8iqakiIHLFFSKbN9f41oUL7W3vvy/idosMHy4SGeGW7N5niLRqJZKfX/mGnA0iK+8T+bCDyBRE3ksQWTJWZO8cEVe533cU3nCrFBMjfxpx4LD2lPXuLdtPP11Oe/11IT1dTHq6DH7jDXl1+XI5WFRUJf+8eSIxMSInnGB/x7PP1vini+Tm2psefzxwnssvt+VQ7v+3hYW337Z2L1wYbksURUREgGUSoF4Ne8VeX6HBCoSIreweeECkUSOR6GiRv/zFCsdhKCsTadZMZOxYkaeftn/N6eNn2si//x34RrdLJGORyLd/FJmaZMXigxNFVt4rkvtT5bzLlomAXGdele3bD2NQSorI+PEiIrLlwAF5+MsvpesLLwjp6RL78MMy/L33ZMa6dVJcVibLl4skJor06iWyf7/IpZeKRESIfPbZYX+2h8REkVtu8X+toMCW5403HsEDQ0BBgUjjxiLjxoXbEkURERWIhsOePSLXX29ryiZNRB57TKSwsNpbrr7a1pNRUSLDLnWJu3dvkc6dRUpLa/bO8iKR7e+JzL9A5H8RVixmnymyeZJIaZ6I2y2lnbrKfHO2TJxY3XPKrd333Vcp2e12y/e7d8vNn38uKU8+KaSnS+Pb/yPxScXSKrVMdux0i4hIXp7IKafYunPNmpqZLl26iIwY4f/a9On28547t4YPCyHjx4skJIjk5Ul+vsiVV2qDQgkfKhANjbVrRS6+2P552rYVeeONgN0kU6fabJ07ixS84nRfvPNO7d5bsEtkzWMiH3f1dEF9M07kjmvFhZFOsbskIyPAvXv22Hf/618BH1/mcsn/lmyVxJQ8oVGBMPEFaffMM3L3nDmyYNs22bC5RFJSRDp2rFEDSuSss0TOPNP/tSuvFElOts2so42vvrJlNXmyXHedjQ4cGG6jlOMVFYiGyoIFIgMG2D9Tnz4is2dXyVJQYLuY1qwosTVr374iLlfd3ut2i2Qutl1Q7zUW+SciIJ/1ukCeeGCH/3uWL7d2zpgR8LHZ2fZnJCSILPy6VKasXi0Xvv22RD74oJCeLpEPPijd7p4hkdHl0q1fnmzYd0DcbndgO0eNssroS3GxbYo43V1HHW63SOfOktHzbAGRHj1s0S1dGm7DlOMRFYiGjMtlWwQdOtg/19ChIitXVs33wgv2+qxZ9fv+snyRrW+KdG0s0gFxvWWk7IshItv+J1Lm1f316af2/d984/cxRUUigwbZrrAvvqh8LauwUGZu2iT3zZsn50yeLLEjPxQQ4ZQV0vKJJ+XSd9+VJ77+Wr76+Wcp9O46u/VWO87gKyKffGJtOaIBjdBy8LaHRUAu6b1N9u+33YTXXBNuq5TjERWIY4HiYpGnnrKj0sbY2mTnTnstL0+kZUuRwYOrVpb1hTMK/sKQCZI9ub3tgpqaJPLdDSK/fCvyysv2c9q2rcqt5eUiw4bZy1OmHP5VZS6XXHdrnoBI/2tWS+fnnhPS04X0dIl+6CEZ+OqrcvPnn8vK226zD83OrvyAceNEkpJESkrq5afXN2VlIsMG/CwujGTd/KCIiEyYIBIbK5KZGWbjlOMOFYhjiQMHRO64w9YmcXEi99wjcued1f7vvV7Ys0fEGJnc8X5p08YlJTvmiyweI/JuvBWLEbHWhm9vF9kxXaRgp4jbLW63HXc/0mmsLpedpWqMbRBk5OfLh+vXy11z5shZ//2vxD3yiFzpqM6gu+6SMTNmyNurVknGwYNWRMeMCV5Z1JH0dFsee3qcI9Kpk4jbLWvX2rTHHgu3dcrxhgrEscj27SJXXWX/hCBy2WXBf+e550pBm5ME3PLKK05aaY7Ilskiw3uKJEaK/C/KCsYURKa3kvUvXyz3XvqwvP7ILJHirCN6XUGBSP/+tvtl9erK10rLy2WDM0L/4P33S/Ljjwvp6XLemDEiIJMffFDmb90qxUfZIPVXX9nJXldfLSJvvmn/dosWiYjIOeeItGt3dI6rK8cu1QmEOutr6KxYAa+8AnfdBR07BvddkybB+PFcc/JSlpSlsWGDl1eNYcOsZ9rVy+HgKshaysZvvkf2L6V7m42eZyR2huQ0SB4IzdOgeT+IahTwlbt3Q1oaxMTA0qXQ0tun74YNcPLJ8PbbuEaP5od9+4i44Qa6z55Nyp13kh8ZSUJ0NIM7dGBo586c37kzXZOTKzsYDCEHD1qP4zEx9s/WJLLAOjUcMQJef50PPrDF+MEHdgsJRQkF1TnrU4FQak52NqSk8NOQG+k681nee8/WbYDdACkhwfqVAqZOhVGj4OKLYfo7OUTlLoespXDge3ss3GXvM5GQ1BNOOANangUtB0F8ZU+wy5bBmWdCv34wf77dSgOA3FxISoInn7SeXV0u64Lj7LPJmzyZBdu3M3vLFr7YsoXNB6zDwnZJSZzfuTNDO3fm3I4daRYfH4KCs828ESPgww9hyRIregCMH28La98+ymMT6NTJOn2thf9GRakVKhBK/TFsGLJkCT2a7CY+MZLlyx0/eB062Fr8rbeYNw8uvNB61549G/zWwUX7IMsRi6zvYP83UO5sptO4qxWKloOsaCScyNSpds+dsWPtxmzGYGvdxET405/g6aft9q2DB9sK94orKr1u68GDzN6yhdlbtjBv2zZyS0qIMIaBqakM7dSJfq1b0zU5mU7NmhEbFVXvxfbaa3DddfCPf9jG3iG++spuHDV5Mowdy9//DvfeC+vW2caRogQbFQil/pg+HS6/nFm3zeHCp85j1iw4f6hAXBzcfDMrRj3BoEFWLxYtgmbNavhcdzkcWAGZX9rwy9dQ5jj7S+gIKYP46JtB3PLYIG64rQN33eV0E3XpAgMGwDvvwE032Zr4l1+scASg3O3mu127DrUuvt+zx7orx27H2qFpU7omJ9O1eXN7dMKJSUlE1KJ7av166N8fTj/dOpeN8HaRKQJdu9p9IhYuJDPTeie//np44YUjfpWiHDEqEEr9UVQErVrhumw47edOoksXWDD9ACQns/+ep+j5+l+Jj4fFi22dV2vcLshe7RGMzEVQaruJdma1hZaDOLH/ILjmFTuGsWCBrVlPPRVmzDiiV+WWlLBx/342ZWWxKSuLjc5xU1YWBWWePb/joqLo4iMaFaFFI//jKMXF8Otf27GU1asDbLb3j3/APffYMZyuXRk71o5D7N5tt7JWlGBSnUDUf1u68osvAJ4DIoHXROQfAfINB94H0sTuSZ1ccQ68ISITg2mncgTEx8OwYUTOmM6d9/6bm++KY9UX++gDPPxaa1xu+7/kOokDQEQkNO9rQ/dbQNyQs47S3V+ycdqX9I6eA0ungAvYFgkvnWn3hhiUCjnroXEXiKjZ590kNpa01FTSfIwWEfbm5x8Si4qwJjOTjzZupNzLZXfLhAT6pKTQ2yuc3KIFd98dxapV8Mkn1ezEOm4c3HcfvP46PP44EyfCW2/ZMGFC7YpPUeqDoLUgjDGRwCZgCLAL+B4YLSLrfPI1BmYCMcBERyASgL5AL6BXTQRCWxAhZM4cGDqUorenk3rTMK7rNI/Hl5/HhXELePDLwUHfvG3vXkhLE05quYkvuk8g9sMv4fwk+CQL/gM0AiLj7OB30z7QtDc0622Pscn1YkO528327Gzb4ti/nx8zM1mdkcGazExKXC4AIn7qinvKaLpetJk/3LOPPq1a0TslhdaJiVVnUl12mR293rkTYmIYOBDy82Ht2qNjryPl2CVcLYiBwGYR2eoY8S7we2CdT76HgceBOyoSRKQA+NoYc1IQ7VNqy9lnQ0oK8TOmcNNNw9jykN20594XW4dkZ8/WreGTTwxnnNGNV/cPZWLRPPg2Gs6/EIb/HQ6utt1T2atg9yewdZLn5vhURzD6eI6Nu9a4tVFBVEQEJzVvzknNm3NRly6H0svdbn7KyuLLNVn89ZlORLc7SOGgz7h73sFDeZLj461YtGxJ75QU+rRqxSnjxxPx4Yfw6acwbBgTJtiGxYIFcM45dS0xRakdwRSIVGCn1/kuoNKu8caYfsCJIjLTGHMHSsMgKspOKXr5ZW5+Mpt339oL2+CMKwL1odQ/ffvaLpgZw9swEey+pZdfAc1OscGbogwrFtmrHfFYBRlzwe2ML0TEQmJHSOjgc3Tisck1/m98VEQE3ZJP4C9/PwFK4dtZMZx88l84WFTEj5mZrNq3j9UZGazOzOTl5cspKi8H4Lz27ZnVpg2Rr70Gw4YxciTcdhu8+KIKhBI+gjoGUR3GmAjgaWBcHZ5xPXA9QLt27erHMKVmXHUVPP88zb/8gD9fuhdebgSNG4fUhGHDoPgPrWESuIhkbuwlDHH7zBICiE+B+KHQeqgnzVUKeRvtor7sHyF/CxRst9NuncHwQ0QleoQjoSMkVhydtJikStn/+U+7HOSVVzxTVZvFx3NW+/ac1b69xwS3my0HDzJr82bumjuX53v04JZZszA7dhDXrh3XXQdPPAE7dtjdShUl1ARzDOI3QLqInO+c3wMgIo8550nAFsCZ/E4r4ABwiYgsc/KMAwboGMRRiIidYtqxI7RoYZc5b9kSejPWrcf07MGi2CEMKplN9+72f95XX21n3taKslzI3w4F2yB/mxWO/G2e84r1GhXENLOD4k26syu3G7c+0J3Uk7vxzKsnYaJi/b2hCqszMrj5pZeY99BDfHPttZz22mvs2GHo1AnuvhsefbSWv0VRDkNYprkaY6Kwg9TnAruxg9RXisjaAPkXArdXiIOTNg4ViKOX+++3NVfXrpCcDF9/HXobCguhUyfKn36eqYzgqaesG4uWLWHiRLjxRqtf9YaIbWFUiEaFcORuwp2zgYji3Z68JgISOkGTbtCke+Vj7AlVuq1yS0rYeuqpNN+2jVteeolJw4YxblQcixfbsetaC56iVEPY1kEYYy4CnsVOc50kIo8aYx7COof62CfvQrwEwhizHWiCnd2UDQz1nQHljQpEGFi/Hnr0sPHLL4dp08Jjh8ihylYEFi6Ep56CmTPtrNxx4+DWW22DJ5iMHQsfTc9j4ceb6Nt5I+RugFznmLcJXMWezDHNoHE3SOpuRaPDGGjUBpk6FTNyJL8dM4b1/ftzR6ur+fOoZN58E8aMCa79yvGJLpRTgke/fvDDD3YV8/PPh9uaSqxbB888A2++CWVl8PvfW5dNp51W/1NHp0yx3VoPPADp6X4yiBsKdlQWjdwNdhykaK9tUZzxHjQ7HVJTyUpL45QhQ8jML6DZ67fTPiWW777T+a5K/VOdQPgO5ynKkXHVVfYYcBVY+OjRA159FX7+2fo3WrQIzjjD+hV8/33r268+2LLFdmWdfrpd7+YXE2EHt9tcAN1vhoEvwXkL4LI98Nt1ENsC5p8Hm5+DsWNJnjOHlZdeytkdO5DRYz5LlxoWLSkL8HBFCQ7aglDqxp490LOn9YV0wQXhtqZaCgqsT7xnnoHNm+34+q23wrXXVnbdVFoKeXnWWWyg4H39668hMxNWrQKvSUpHRlkefDcedkwD9xAYMweeeAL37bfzt88X8/fL0mjadxvffNqC7vU6qKIc72gXkxJc3P7mlh69uFzw8cd2OuqSJdZjeIsWngq/pOTwzzDG+klq0sQ6JHzsMbjoojoaJgIbnoaVd8EjMVCaApu2gjFcMiaHT95NoNGdL/L66PMY1atXHV+mKBYVCEUJwDff2G6okhJPhe8vNG5c+TwhIYguMDIWwiOXwIt5MPVBuOJ+1q2zDbX2w5fx869mMiEtjaeGDg2Ka3Ll+EIFQlEaGvt/gvYnQ18XvHAH9Pk75w6JYvNmYfh/5vLM0iWktWnD1CuuoEPTpuG2VmnA6CC1ojQ0WnSBseNhWSQsexLmD+GvEzLZscNwVskQZowYwcasLPq9/DIzN20Kt7XKMYq2IBTlaOWHH+w04vQx0H0aEpPMpf98n8JGv2bOHNhy4ACXT5vGyn376NWyJW2bNOHEJk1o6xUqzhvH1mxFt3L8oV1MitJQ6d/fjqrP/y98fTmuvJ3c9Maz3PTcjZzcw1BUVsYTixfzw7597MrNZVduLhkFBVUe0yQ2topoeIcmsbHEREYeCrGRkURFRFR1S64cc4RtwyBFUerIH/8If/4zbHXBBcsoX3A1/752At/N/g66vkR8dCMeGDy40i0l5eXsycs7JBiHQl4eO3Ny+DEjg335+dTkv4beglFJQKKiKp3HR0URHx1NfFQUjbyP0dGHzr3jvtcSYmJIiI4mMSaGuKgoFaajBG1BKMrRTE6OXYQ4Zgy8/DKIm+kPPcJlndORpr2JHDwDEjsd8WPLXC725ucfEo/80lJKXa5DoaS83BP3Svc9Lykvp8Tlori8nKKyMorKyyksK6OorIzCsrJDmycdCRHGHBKLhJgYe/Q5T4yOrnQtLiqK2KioQ0JWIWD+zv1di42MPG5FSbuYFKUhM24cTJ9ut9JLTGTpUnjg+s/54PariIsVOG0KpNZ1EUZwcItQ7CUaRX7iFSG/tJSC0lJ7dM6rxH3Sip39NOpKpDE0iY0lKS7OHmNjD50fivtLc+KRERGUu921DnFRUTSNi6sUmsTGEhWC9UUqEIrSkFm82PoImTTJLvsG0tKgWfRWvrhvOCZ7ld0ZL76NE1KhURuv8zbW11NEZJh/SP3jcrvJLy2lxKvV4xsvDXDuHc8vLSW3pIQcJ+SWlJBTXOxJKy6mzGsP8lDROCaminAkxcXRNDa2Ulr3Fi04vZabhqhAKEpDRsSukmva1C79xroMGTcOFswtYnCLR+3GR0V7bCjOAN8RBhMJca2sWDTyEpL4NhDf2vqCimlmQ3TSMSkmdUFEKHG5KolGhYjklJTgFiEqIqJWIdIYisvLyS4u9h9KSvym5xQXH/orj+rVi3eGD6/Vb1OBUJSGztNP252Q1qyBnj0pLoa2beGss2DGDJ+87nIrEhWCUbjbE/cOJVkBXmYguolHMPyGphAdIP0I9/dWaodbhDxHPCIjImjbpEmtnqMCoSgNnV9+gdRUmDDBehsE7rnHbkm6bVsttyR1FUPRPo9YlB50wgGvuJ/gPoyzqqjGVUWjOrGJbgqRFes0nIFiYzxxjJdfkwBpUY3s1rBG1/4eKSoQinIsMGIEzJtnPejGxvLzz4RnS9LyosDiUZYd4JqT7ioMomHGdo/FJNljdJLT0vGJH7re1BOPjIeIaDBR9ugdN5FBdLwVflQgFOVYYM4cGDrUulYfNQqASy+lYW1J6irxiIW3mLjLALHjLYBnDKUmaQLlhfa5ZTk2BIpLLQeaD4mFPwGJclowCbYVUxGiveLe1yqlV1xzQmR8yMVIBUJRjgXcbujc2Ya5cwF7GDLEhrQ06NbNhq5drRvyhkBODjRqBNHRQX6RCJQXOMLkIx6uYpAyO37jLgNxjt7xgNfL7P3l+VCWb4/ewXur2cNiPGITmeAlPBXn3qGRJy2pJ7Q5v1bFEraV1MaYC4DnsHtSvyYi/wiQbzjwPpDmtSf1PcB4wAX8RUS+CKatinLUExEB48fD3/4GW7dCp06ce67d7XXWLJg/v/IueSecYIXCWzS6dbPdUuF0zSRix9o/+cSG776D5s3ttuajR8OZZwZpexFj7P/eoxOhUdsgvCAAbhe4CqqKR1m+k55nhau8wHbBVcR900p3V01zl9p3tB9da4GojqC1IIwxkcAmYAiwC/geGC0i63zyNQZmAjHARBFZZozpAbwDDATaAHOBriIScFmmtiCU44Jdu+y2dX4GHkpL7YD1xo2waVPlY0aGJ19EhN1Nz1s0evWCPn3svhfBoKQEFi60gvDpp3YbWIABA+DCC+0Ofx99BIWF0KYNjBxpxWLAgGO6+7/uuMutUIAdT6kFYeliMsb8BkgXkfOd83sAROQxn3zPAnOAO4DbHYGolNcY84XzrG8CvU8FQjlu+N3vYMUK2LEDarhhUE6OFQtf4di0yVbKFZx0EvTtC6ecYkPfvtCqVe0q6cxMmDnTCsLs2ZCfD/Hxtjvsd7+D3/7WikEFBQVWQN59Fz7/3Ape5852uGX0aLsURKl/wtXFlArs9DrfBZzqY1g/4EQRmWmMucPn3m997k31fYEx5nrgeoB2tVxFqCgNjuuus6PTn30Gl1xSo1uSkuwYRVpa5XS3G3bvhtWrYeVK62F8xQqYNs2Tp2XLqqJx0kkQ6bOWzl/XkYidnXvVVXDxxXDOOVYk/JGQYMVg1CjIzrbrO959127n+uijtpUzerS93unI3U8ptSCYLYjLgQtE5I/O+RjgVBGZ6JxHAPOBcSKy3RizEE8L4kXgWxF528n7OvC5iLwf6H3aglCOG8rK7MKHtDS7uXYQyMmxovHDDx7hWLvWvhrsoHKfPlYwevWCdeuqdh1dfLENp5xSt26ijAx4/307eWvxYps2cKAVipEjK7dClCPnqOxiMsYkAVuAfOeWVsAB4BLsuIV2MSlKIP7v/+Dxx203U2qVxnVQKC2F9esri8bKlZCb6+k6uvhiuOii4FXaO3bAe+9ZsfjhBys8gwZBjx52bMVfiIw8/LXYWNuCqQiJiZXPK0J86GehBp1wCUQUdpD6XGA3dpD6ShFZGyD/QjwtiJ7A//AMUs8DuuggtaI4bN4MXbrAI4/AvfeGzQwRuwbjhBMCdx0Fi40bbRfUtGm2leF2Vw4uV9W0ulZ3xtjWk6+QxMfbEBdnQ0XcX5q/6xERHntrc+zYEc49t7a/KUzrIIwxFwHPYqe5ThKRR40xDwHLRORjn7wLcQTCOb8X+ANQDtwiIp9X9y4VCOW445xzYPt2KxYhcAt9LCDOGjvfCrakxA6iFxRUHwLlKSqyobjYBt94sJebjRxpxbI26EI5RTkWeecduPJKu8L6vPPCbY0SABGNAle2AAAJNElEQVQ7duNPOLwFxLvL60iPCQl2LUlt0C1HFeVY5LLLbK3w2msqEEcxxkBMjA21dLgaNlQgFKWhEhdntyJ96SXr7TU+HvLyqobcXP/pFSE/3y526NnTTknq2dPOY63hGgvlKMDlqjrvuB7QLiZFacj8+CP07n1k9zRuXDkkJNgV2lu3ejrLY2I8S6x79vSETp1qVxHl5NjRbH9h925rR8eO9vnex3btrC1KZbKy4MsvYcECG8480/5HoRZoF5OiHKv86le2YqioZJs0qSoAvmIQaEC7sNDOY1271hOWLLFjHRXExcHJJ1cWjR49bCe7b8W/a5cnnpdX+V0REdC6NZx4ohWhvDw7b/XDDz2LLSrytW1bVTgqjikpx968U39kZ8OiRR5BWL3ainmjRnY7Wt8VkPWEtiAURamevDyPcKxZ4xGPXbsC35OSYit/39C2rT22bu3ffavLZfe72LrVOpbyPe7dWzl/fLwViubN7WKGihAXV/Pzivmm8fG2wvUOFWlxcaEVorw8+OorjyD88IOdchUXB6edBmefbUNaWp1bWDqLSVGU+icnxy6hXrfOVlwVIpCaGrxuoaIiO7XXVzhycuxc1ZISOzWoIu57Xhe8BcNXRCpab0cSvMuooMAuE68QhGXLrFhGR8NvfuMRhFNPrfeNP1QgFEVRvOeb+gpIcbHtYisstCJUEa/JeWGhHejPzfUEV8A1vR5iY61QJCba1lhZmZ0YkJbmEYTTTrMiFER0DEJRFMV7vmkwEbEi4i0Y/kJOjifetq0VhDPOsIJxlKACoSiKUp9U+ONo1MhOH27A6Pp8RVEUxS8qEIqiKIpfVCAURVEUv6hAKIqiKH5RgVAURVH8ogKhKIqi+EUFQlEURfGLCoSiKIril2PG1YYx5hfg53DbUQ0tgP3hNqIa1L66ofbVDbWvbtTFvvYicoK/C8eMQBztGGOWBfJ3cjSg9tUNta9uqH11I1j2aReToiiK4hcVCEVRFMUvKhCh45VwG3AY1L66ofbVDbWvbgTFPh2DUBRFUfyiLQhFURTFLyoQiqIoil9UIOoJY8yJxpgFxph1xpi1xpib/eQZbIzJMcasdML9IbZxuzHmR+fdVfZnNZbnjTGbjTGrjTH9QmhbN69yWWmMyTXG3OKTJ+TlZ4yZZIzJNMas8UprboyZY4z5yTk2C3DvNU6en4wx14TQvieNMRucv+EHxpimAe6t9nsIon3pxpjdXn/HiwLce4ExZqPzPd4dQvve87JtuzFmZYB7Q1F+fuuVkH2DIqKhHgLQGujnxBsDm4AePnkGA5+G0cbtQItqrl8EfA4Y4NfAd2GyMxLYh13AE9byA84C+gFrvNKeAO524ncDj/u5rzmw1Tk2c+LNQmTfUCDKiT/uz76afA9BtC8duL0G38AWoBMQA6zy/fcULPt8rj8F3B/G8vNbr4TqG9QWRD0hIntFZIUTzwPWA6nhteqI+T3wpli+BZoaY1qHwY5zgS0iEvaV8SKyCDjgk/x7YLITnwxc6ufW84E5InJARA4Cc4ALQmGfiMwWkXLn9FugbX2/t6YEKL+aMBDYLCJbRaQUeBdb7vVKdfYZYwwwAninvt9bU6qpV0LyDapABAFjTAegL/Cdn8u/McasMsZ8bozpGVLDQIDZxpjlxpjr/VxPBXZ6ne8iPCI3isD/KMNZfhWkiMheJ74PSPGT52gpyz9gW4X+ONz3EEwmOl1gkwJ0jxwN5XcmkCEiPwW4HtLy86lXQvINqkDUM8aYRGA6cIuI5PpcXoHtNukDvAB8GGLzzhCRfsCFwARjzFkhfv9hMcbEAJcA0/xcDnf5VUFsW/6onCtujLkXKAemBMgSru/hJaAzcAqwF9uNczQymupbDyErv+rqlWB+gyoQ9YgxJhr7R5wiIjN8r4tIrojkO/HPgGhjTItQ2Sciu51jJvABthnvzW7gRK/ztk5aKLkQWCEiGb4Xwl1+XmRUdL05x0w/ecJalsaYccDvgKucCqQKNfgegoKIZIiIS0TcwKsB3hvu8osChgHvBcoTqvILUK+E5BtUgagnnP7K14H1IvJ0gDytnHwYYwZiyz8rRPYlGGMaV8SxA5lrfLJ9DIx1ZjP9GsjxasaGioD/awtn+fnwMVAxI+Qa4CM/eb4AhhpjmjldKEOdtKBjjLkAuBO4REQKA+SpyfcQLPu8x7UuC/De74EuxpiOTqtyFLbcQ8V5wAYR2eXvYqjKr5p6JTTfYDBH4I+nAJyBbeatBlY64SLgBuAGJ89EYC12Rsa3wGkhtK+T895Vjg33Oune9hngX9jZIz8CA0JchgnYCj/JKy2s5YcVq71AGbYPdzyQDMwDfgLmAs2dvAOA17zu/QOw2QnXhtC+zdi+54rv8D9O3jbAZ9V9DyGy7y3n+1qNreha+9rnnF+EnbWzJZT2OelvVHx3XnnDUX6B6pWQfIPqakNRFEXxi3YxKYqiKH5RgVAURVH8ogKhKIqi+EUFQlEURfGLCoSiKIriFxUIRTkMxhiXqexptt48ixpjOnh7ElWUo4mocBugKA2AIhE5JdxGKEqo0RaEotQSZz+AJ5w9AZYaY05y0jsYY+Y7zujmGWPaOekpxu7PsMoJpzmPijTGvOr4+59tjIl38v/F2QdgtTHm3TD9TOU4RgVCUQ5PvE8X00ivazki8ivgReBZJ+0FYLKI9MY6ynveSX8e+FKss8F+2BW4AF2Af4lITyAbGO6k3w30dZ5zQ7B+nKIEQldSK8phMMbki0iin/TtwDkistVxqLZPRJKNMfux7iPKnPS9ItLCGPML0FZESrye0QHrs7+Lc34XEC0ijxhjZgH5WK+1H4rjqFBRQoW2IBSlbkiA+JFQ4hV34Rkb/C3WN1Y/4HvHw6iihAwVCEWpGyO9jt848SVY76MAVwFfOfF5wI0AxphIY0xSoIcaYyKAE0VkAXAXkARUacUoSjDR/5EoyuGJN5U3rp8lIhVTXZsZY1ZjWwGjnbSbgP8aY+4AfgGuddJvBl4xxozHthRuxHoS9Uck8LYjIgZ4XkSy6+0XKUoN0DEIRaklzhjEABHZH25bFCUYaBeToiiK4hdtQSiKoih+0RaEoiiK4hcVCEVRFMUvKhCKoiiKX1QgFEVRFL+oQCiKoih++X9w4ynyvljDWgAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "u3fptpSJnF1L"
      },
      "source": [
        "From the plots, we can saw a general, consistent trend between the two models:\n",
        "\n",
        "1. The static version is more unstable compared to the dynamic model. There are more (and deeper) local minima seems to exist in the loss landscape of the static version, shown by the more \"jumpy\" validation curve. This shows that dynamically updating the graph helps to stabilize the optimization process.\n",
        "\n",
        "2. But interestingly, the static version final performance, when the models reached convergence, is consistently a little bit better than the dynamic one for both training and validation metric (lower loss, higher accuracy and AUC).\n",
        "\n",
        "3. The training time for the static version (211s/epoch) is lowered than the dynamic version (311s/epoch). This is to be expected since the static version does not need to recalculate the nearest neighboring particles for every EdgeConv block in the network. The difference in time may differs from one hardware to another, but since the difference is quite significant, it is to be expected that the training time for the static model will still be significantly faster than the dynamic model.\n",
        "\n",
        "From this results, one question arises, does the \"dynamic\" aspect of the ParticleNet is important in improving the model's performance? Considering it takes a more significant ammount of training time (and computational power) without a clear improvement in the performance, it is probably wise to reconsider the \"static\" option.\n",
        "\n",
        "The ones that are significantly improving the performance compared to other approaches are probably the EdgeConv operation itself (without updating the graph dynamically) and the particle cloud representation."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WyVDfBC-qG0p"
      },
      "source": [
        "Let's evaluate both models on testing set."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "64diHfXfcvRH",
        "outputId": "93a5a824-d22a-4b7c-8f78-0b360bb6b022"
      },
      "source": [
        "# test dataset evaluation for DGCNN\n",
        "\n",
        "loss, acc, auc = model_dynamic.evaluate(test_dict, y_test)\n",
        "print(\"Evaluation Metric of Dynamic Graph CNN Model on Test Dataset\")\n",
        "print(\"Loss:\", loss)\n",
        "print(\"Accuracy:\", acc)\n",
        "print(\"AUC:\", auc)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2813/2813 [==============================] - 41s 14ms/step - loss: 0.4032 - accuracy: 0.8240 - auc_1: 0.8987\n",
            "Evaluation Metric of Dynamic Graph CNN Model on Test Dataset\n",
            "Loss: 0.4032370150089264\n",
            "Accuracy: 0.8239666819572449\n",
            "AUC: 0.8986521363258362\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uyRgRwo1c0lY",
        "outputId": "d1c6a45a-5790-4d70-ba5e-b1e8489351c8"
      },
      "source": [
        "# test dataset evaluation for SGCNN\n",
        "\n",
        "loss, acc, auc = model_static.evaluate(test_dict, y_test)\n",
        "print(\"Evaluation Metric of Static Graph CNN Model on Test Dataset\")\n",
        "print(\"Loss:\", loss)\n",
        "print(\"Accuracy:\", acc)\n",
        "print(\"AUC:\", auc)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2813/2813 [==============================] - 29s 10ms/step - loss: 0.4019 - accuracy: 0.8243 - auc: 0.8994\n",
            "Evaluation Metric of Static Graph CNN Model on Test Dataset\n",
            "Loss: 0.40190747380256653\n",
            "Accuracy: 0.8243444561958313\n",
            "AUC: 0.899357259273529\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XsotFlT5qlEq"
      },
      "source": [
        "The table below summarizes all the best results obtained from both models in this project (when convergence is reach) and the original paper's results (reference [2]). Note that the model used in the original paper (ParticleNet-Lite w/ PID) is the same as the dynamic model used in this project, the difference is in the total number of dataset used."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P6kX3o4MrvVE",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "outputId": "98285cad-73af-49a6-9c54-50261a901a3a"
      },
      "source": [
        "import pandas as pd\n",
        "\n",
        "summary = {'Accuracy': [0.825, 0.823, 0.829, 0.827, 0.824, 0.824, 0.835],\n",
        "           'AUC': [0.8998, 0.8987, 0.9030, 0.9014, 0.8994, 0.8987, 0.9079]}\n",
        "row_label = [\n",
        "             'Train (Static)', 'Train (Dynamic)',\n",
        "             'Validation (Static)', 'Validation (Dynamic)',\n",
        "             'Test (Static)', 'Test (Dynamic)',\n",
        "             'Reported in the Paper[2]'\n",
        "             ]\n",
        "\n",
        "pd.DataFrame(data=summary, index=row_label)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Accuracy</th>\n",
              "      <th>AUC</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>Train (Static)</th>\n",
              "      <td>0.825</td>\n",
              "      <td>0.8998</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Train (Dynamic)</th>\n",
              "      <td>0.823</td>\n",
              "      <td>0.8987</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Validation (Static)</th>\n",
              "      <td>0.829</td>\n",
              "      <td>0.9030</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Validation (Dynamic)</th>\n",
              "      <td>0.827</td>\n",
              "      <td>0.9014</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Test (Static)</th>\n",
              "      <td>0.824</td>\n",
              "      <td>0.8994</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Test (Dynamic)</th>\n",
              "      <td>0.824</td>\n",
              "      <td>0.8987</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Reported in the Paper[2]</th>\n",
              "      <td>0.835</td>\n",
              "      <td>0.9079</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                          Accuracy     AUC\n",
              "Train (Static)               0.825  0.8998\n",
              "Train (Dynamic)              0.823  0.8987\n",
              "Validation (Static)          0.829  0.9030\n",
              "Validation (Dynamic)         0.827  0.9014\n",
              "Test (Static)                0.824  0.8994\n",
              "Test (Dynamic)               0.824  0.8987\n",
              "Reported in the Paper[2]     0.835  0.9079"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 2
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "73FX6gGzaaK7"
      },
      "source": [
        "With the test set results now included, it is clear that the dynamic model does not offer a clear improvement in performance, both accuracy and AUC. The numerical results obtained here may include some uncertainty and this experiment needs to be carried out several times in the future to calculate the mean and the standard deviation. But at least this results show a clear indication that even without dynamically updating the graph between EdgeConv, the model still performed as well."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3tO6bVUdb_e5"
      },
      "source": [
        "It is also interesting to note that the total samples used in the dataset mattered in the model's performance. The original paper used 2000k dataset, more than twice the dataset used in this project, and the performance of the original paper's model is indeed better (to be expected as more data usually means better model)."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Q7Ay0K6Dd_RL"
      },
      "source": [
        "## Conclusion"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DpN-fIEIeNuO"
      },
      "source": [
        "In this project, two models based on reference [2] are trained to do quark-gluon tagging, using a much lower number of training samples. The first model is the dynamic model, the same model as the ParticleNet-Lite w/ PID from reference [2]. The second model is the static model, a similar architecture as the ParticleNet-Lite w/ PID but without dynamically updating the graph through the networks.\n",
        "\n",
        "From the numerical results, we may conclude that the \"dynamic\" aspect of the first model does not show any clear improvement in the model's performance, at least for the lite version of ParticleNet. Considering that the first model significantly needs more training time, it is recommended to reconsider the second model (Static Graph CNN) as it is able to give a similar performance with lower training time (and computational power)."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "klXUsW6YBLLC"
      },
      "source": [
        "## References\n",
        "\n",
        "[1] [P. Komiske, E. Metodiev, and J. Thaler, “Pythia8 Quark and Gluon Jets for Energy Flow,” May 2019.](https://zenodo.org/record/3164691#.Xk1VwS2B1QI) <br>\n",
        "[2] [H. Qu and L. Gouskos, “Jet tagging via particle clouds,” Phys. Rev. D, vol. 101, no. 5, p. 056019, Mar. 2020.](https://journals.aps.org/prd/abstract/10.1103/PhysRevD.101.056019) <br>\n",
        "[3] [Y. Wang, Y. Sun, Z. Liu, S. E. Sarma, M. M. Bronstein, and J. M. Solomon, “Dynamic Graph CNN for Learning on Point Clouds,” ACM Trans. Graph., vol. 38, no. 5, p. Article 146, 2019.](https://dl.acm.org/doi/10.1145/3326362)"
      ]
    }
  ]
}