4_qumode_classifier.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "4d6aa70b",
   "metadata": {},
   "source": [
    "### Author: Sophie Choe"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "97b10033",
   "metadata": {},
   "source": [
    "#### Portland State University, Electrical and Computer Engineering"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "595ccfec",
   "metadata": {},
   "source": [
    "Dependencies: keras-nightly==2.5.0.dev2021032900 PennyLane==0.17.0 StrawberryFields==0.18.0 tensorflow-2.4.0-cp38-cp38-macosx_10_9_x86_64.whl"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c2b1738d",
   "metadata": {},
   "source": [
    "# 4-qumode classifier\n",
    "\n",
    "Classical and Continuous Variable Quantum hybrid network: Classical layers using keras dense and quantum layers using Pennylane"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "417078e3",
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import tensorflow_datasets as tfds\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "\n",
    "import pennylane as qml\n",
    "import numpy as np\n",
    "\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "84a81d5b",
   "metadata": {},
   "source": [
    "## 0. Loading data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3d91f7fa",
   "metadata": {},
   "source": [
    "Normalize pixel values from 0 ~ 255 to 0 ~ 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "9cb00f94",
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist = keras.datasets.mnist\n",
    "\n",
    "# datasets are numpy.ndarrays\n",
    "(X_train, Y_train), (X_test, Y_test) = mnist.load_data()     \n",
    "\n",
    "# normalize the image data\n",
    "X_train, X_test = X_train / 255.0, X_test / 255.0"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "642c2280",
   "metadata": {},
   "source": [
    "One hot encode labels to vectors of size cutoff_dim^(num_qumodes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "849f33ec",
   "metadata": {},
   "outputs": [],
   "source": [
    "def one_hot(labels):  \n",
    "       \n",
    "    depth =  2**4                       # 10 classes + 6 zeros for padding\n",
    "    indices = labels.astype(np.int32)    \n",
    "    one_hot_labels = np.eye(depth)[indices].astype(np.float32) \n",
    "    \n",
    "    return one_hot_labels\n",
    "\n",
    "# one-hot encoded labels, each label of length cutoff dimension**2\n",
    "y_train, y_test = one_hot(Y_train), one_hot(Y_test)\n",
    "\n",
    "# using only 600 samples for training in this experiment\n",
    "n_samples = 600\n",
    "test_samples = 100\n",
    "X_train, X_test, y_train, y_test = X_train[:n_samples], X_test[:test_samples], y_train[:n_samples], y_test[:test_samples]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8185cf3c",
   "metadata": {},
   "source": [
    "## 1. Classical circuit"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "54114c44",
   "metadata": {},
   "source": [
    "Define classical layers using Keras Sequential. Take in 28x28 image matrices, flatten, and output vectors of length 30. 2 hidden layers with ELU activation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "46bb059f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten (Flatten)            (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 392)               307720    \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 196)               77028     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 98)                19306     \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 49)                4851      \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 30)                1500      \n",
      "=================================================================\n",
      "Total params: 410,405\n",
      "Trainable params: 410,405\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "keras.backend.set_floatx('float32')\n",
    "\n",
    "model = keras.models.Sequential([\n",
    "                                 layers.Flatten(input_shape = (28,28)),\n",
    "                                 layers.Dense(392, activation =\"elu\"),\n",
    "                                 layers.Dense(196, activation =\"elu\"),\n",
    "                                 layers.Dense(98),\n",
    "                                 layers.Dense(49, activation =\"elu\"),\n",
    "                                 layers.Dense(30),\n",
    "                                ])\n",
    "\n",
    "# More than a million parameters for the classical circuit\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b3e30830",
   "metadata": {},
   "source": [
    "## 2. Data encoding circuit"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a7c72d96",
   "metadata": {},
   "source": [
    "Encode the output vectors from the classical network into quantum states using the vector entries as parameters of continuous variable gates. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "6a5e27b4",
   "metadata": {},
   "outputs": [],
   "source": [
    "def encode_data(x):\n",
    "    qml.Squeezing(x[0], x[1], wires=0)\n",
    "    qml.Squeezing(x[2], x[3], wires=1)\n",
    "    qml.Squeezing(x[4], x[5], wires=2)\n",
    "    qml.Squeezing(x[6], x[7], wires=3)\n",
    "    \n",
    "    qml.Beamsplitter(x[8], x[9], wires=[0,1])\n",
    "    qml.Beamsplitter(x[10], x[11], wires=[1,2])\n",
    "    qml.Beamsplitter(x[12], x[13], wires=[2,3])\n",
    "    \n",
    "    qml.Rotation(x[14], wires=0)\n",
    "    qml.Rotation(x[15], wires=1)\n",
    "    qml.Rotation(x[16], wires=2)\n",
    "    qml.Rotation(x[17], wires=3)    \n",
    "    \n",
    "    qml.Displacement(x[18], x[19], wires=0)\n",
    "    qml.Displacement(x[20], x[21], wires=1)\n",
    "    qml.Displacement(x[22], x[23], wires=2)\n",
    "    qml.Displacement(x[24], x[25], wires=3) \n",
    "    \n",
    "    qml.Kerr(x[26], wires=0)\n",
    "    qml.Kerr(x[27], wires=1)\n",
    "    qml.Kerr(x[28], wires=2)\n",
    "    qml.Kerr(x[29], wires=3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9eca77da",
   "metadata": {},
   "source": [
    "## 3. Qauntum neural network circuit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "de0894c2",
   "metadata": {},
   "outputs": [],
   "source": [
    "def layer(v):\n",
    "    \n",
    "    # Linear transformation W = Interferemeter, squeezers, interferometer\n",
    "    # Interferometer 1\n",
    "    qml.Beamsplitter(v[0], v[1], wires=[0,1])\n",
    "    qml.Beamsplitter(v[2], v[3], wires=[1,2])\n",
    "    qml.Beamsplitter(v[4], v[5], wires=[2,3])\n",
    "    \n",
    "    qml.Rotation(v[6], wires=0)\n",
    "    qml.Rotation(v[7], wires=1)\n",
    "    qml.Rotation(v[8], wires=2)\n",
    "    qml.Rotation(v[9], wires=3)\n",
    "    \n",
    "    # Squeezers\n",
    "    qml.Squeezing(v[10], 0.0, wires=0)\n",
    "    qml.Squeezing(v[11], 0.0, wires=1)\n",
    "    qml.Squeezing(v[12], 0.0, wires=2)\n",
    "    qml.Squeezing(v[13], 0.0, wires=3) \n",
    "    \n",
    "    # Interferometer 2\n",
    "    qml.Beamsplitter(v[14], v[15], wires=[0,1])\n",
    "    qml.Beamsplitter(v[16], v[17], wires=[1,2])\n",
    "    qml.Beamsplitter(v[18], v[19], wires=[2,3])\n",
    "    \n",
    "    qml.Rotation(v[20], wires=0)\n",
    "    qml.Rotation(v[21], wires=1)\n",
    "    qml.Rotation(v[22], wires=2)\n",
    "    qml.Rotation(v[23], wires=3)\n",
    "    \n",
    "    # Bias addition\n",
    "    qml.Displacement(v[24], 0.0, wires=0)\n",
    "    qml.Displacement(v[25], 0.0, wires=1)\n",
    "    qml.Displacement(v[26], 0.0, wires=2)\n",
    "    qml.Displacement(v[27], 0.0, wires=3)\n",
    "    \n",
    "    # Non-linear activation\n",
    "    qml.Kerr(v[28], wires=0)\n",
    "    qml.Kerr(v[29], wires=1)\n",
    "    qml.Kerr(v[30], wires=2)\n",
    "    qml.Kerr(v[31], wires=3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9f2b136b",
   "metadata": {},
   "source": [
    "## 4. Quantum device"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d07c0173",
   "metadata": {},
   "source": [
    "For the expression of qumodes in Fock basis, choose a \"strawberryfields.fock\" device. Define the number of qumodes and cutoff dimension. Run the data encoding circuit and quantum neural network circuit. The probability measurement method (qml.probs(wires)) returns vectors of size 2^4 = 16 (cutoff_dim^num_modes)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "83dee8c3",
   "metadata": {},
   "outputs": [],
   "source": [
    "num_modes = 4\n",
    "cutoff_dim = 2\n",
    "\n",
    "# select a devide \n",
    "dev = qml.device(\"strawberryfields.fock\", wires=num_modes, cutoff_dim=cutoff_dim) \n",
    "\n",
    "@qml.qnode(dev, interface=\"tf\")\n",
    "def quantum_nn(inputs, var):\n",
    "    # Encode input x into quantum state\n",
    "    encode_data(inputs)\n",
    "\n",
    "    # iterative quantum layers\n",
    "    for v in var:\n",
    "        layer(v)\n",
    "\n",
    "    return qml.probs(wires=[0, 1, 2, 3])  # Measurement"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "96eedc15",
   "metadata": {},
   "source": [
    "## 5. Hybrid circuit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "bb3d72be",
   "metadata": {},
   "outputs": [],
   "source": [
    "weight_shape = {'var': (4,32)}          # 4 layers and 32 parameters per layer, Keras layer will initialize.\n",
    "\n",
    "num_layers = 4\n",
    "qlayer = qml.qnn.KerasLayer(quantum_nn, weight_shape, output_dim = 4)\n",
    "\n",
    "# add to the classical sequential model\n",
    "model.add(qlayer)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "198357cc",
   "metadata": {},
   "source": [
    "## 6. Loss function and optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "1d32cad7",
   "metadata": {},
   "outputs": [],
   "source": [
    "opt = keras.optimizers.SGD(lr = 0.02)\n",
    "model.compile(opt, loss = 'categorical_crossentropy', metrics =['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c433de7a",
   "metadata": {},
   "source": [
    "## 7. Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "44b23d51",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/70\n",
      "10/10 [==============================] - 1382s 140s/step - loss: 2.9951 - accuracy: 0.1697 - val_loss: 2.2072 - val_accuracy: 0.2700\n",
      "Epoch 2/70\n",
      "10/10 [==============================] - 1655s 170s/step - loss: 1.9471 - accuracy: 0.3921 - val_loss: 2.6418 - val_accuracy: 0.2600\n",
      "Epoch 3/70\n",
      "10/10 [==============================] - 1938s 197s/step - loss: 1.9178 - accuracy: 0.4227 - val_loss: 1.6940 - val_accuracy: 0.4700\n",
      "Epoch 4/70\n",
      "10/10 [==============================] - 12172s 1334s/step - loss: 1.5007 - accuracy: 0.5805 - val_loss: 1.7040 - val_accuracy: 0.4900\n",
      "Epoch 5/70\n",
      "10/10 [==============================] - 14883s 1466s/step - loss: 1.4438 - accuracy: 0.5998 - val_loss: 1.3845 - val_accuracy: 0.5900\n",
      "Epoch 6/70\n",
      "10/10 [==============================] - 7353s 661s/step - loss: 1.1971 - accuracy: 0.7134 - val_loss: 1.3443 - val_accuracy: 0.6000\n",
      "Epoch 7/70\n",
      "10/10 [==============================] - 6382s 690s/step - loss: 1.1569 - accuracy: 0.7099 - val_loss: 1.3299 - val_accuracy: 0.6100\n",
      "Epoch 8/70\n",
      "10/10 [==============================] - 5532s 518s/step - loss: 1.0633 - accuracy: 0.7327 - val_loss: 1.3823 - val_accuracy: 0.5900\n",
      "Epoch 9/70\n",
      "10/10 [==============================] - 1820s 183s/step - loss: 1.0054 - accuracy: 0.7659 - val_loss: 1.2668 - val_accuracy: 0.6600\n",
      "Epoch 10/70\n",
      "10/10 [==============================] - 1814s 182s/step - loss: 0.9874 - accuracy: 0.7804 - val_loss: 1.1943 - val_accuracy: 0.6200\n",
      "Epoch 11/70\n",
      "10/10 [==============================] - 4285s 409s/step - loss: 0.9234 - accuracy: 0.7911 - val_loss: 1.2333 - val_accuracy: 0.6200\n",
      "Epoch 12/70\n",
      "10/10 [==============================] - 1854s 176s/step - loss: 0.8662 - accuracy: 0.7978 - val_loss: 1.0955 - val_accuracy: 0.6600\n",
      "Epoch 13/70\n",
      "10/10 [==============================] - 2911s 303s/step - loss: 0.8484 - accuracy: 0.8037 - val_loss: 1.1831 - val_accuracy: 0.6400\n",
      "Epoch 14/70\n",
      "10/10 [==============================] - 1737s 171s/step - loss: 0.8317 - accuracy: 0.8091 - val_loss: 1.2126 - val_accuracy: 0.6400\n",
      "Epoch 15/70\n",
      "10/10 [==============================] - 4082s 432s/step - loss: 0.7482 - accuracy: 0.8082 - val_loss: 1.0863 - val_accuracy: 0.6900\n",
      "Epoch 16/70\n",
      "10/10 [==============================] - 1743s 168s/step - loss: 0.7280 - accuracy: 0.8267 - val_loss: 1.2221 - val_accuracy: 0.6400\n",
      "Epoch 17/70\n",
      "10/10 [==============================] - 4198s 448s/step - loss: 0.6950 - accuracy: 0.8291 - val_loss: 1.2001 - val_accuracy: 0.6500\n",
      "Epoch 18/70\n",
      "10/10 [==============================] - 1846s 184s/step - loss: 0.7085 - accuracy: 0.8383 - val_loss: 1.0801 - val_accuracy: 0.7000\n",
      "Epoch 19/70\n",
      "10/10 [==============================] - 2047s 208s/step - loss: 0.5855 - accuracy: 0.8769 - val_loss: 1.0720 - val_accuracy: 0.6600\n",
      "Epoch 20/70\n",
      "10/10 [==============================] - 4410s 464s/step - loss: 0.5897 - accuracy: 0.8683 - val_loss: 1.0643 - val_accuracy: 0.6800\n",
      "Epoch 21/70\n",
      "10/10 [==============================] - 1908s 191s/step - loss: 0.5772 - accuracy: 0.8704 - val_loss: 0.9931 - val_accuracy: 0.7400\n",
      "Epoch 22/70\n",
      "10/10 [==============================] - 4217s 447s/step - loss: 0.5320 - accuracy: 0.8956 - val_loss: 1.2893 - val_accuracy: 0.6000\n",
      "Epoch 23/70\n",
      "10/10 [==============================] - 15053s 1541s/step - loss: 0.6110 - accuracy: 0.8575 - val_loss: 1.0679 - val_accuracy: 0.7200\n",
      "Epoch 24/70\n",
      "10/10 [==============================] - 21399s 2223s/step - loss: 0.5024 - accuracy: 0.8974 - val_loss: 0.9927 - val_accuracy: 0.6800\n",
      "Epoch 25/70\n",
      "10/10 [==============================] - 3445s 228s/step - loss: 0.4774 - accuracy: 0.9094 - val_loss: 1.1123 - val_accuracy: 0.6900\n",
      "Epoch 26/70\n",
      "10/10 [==============================] - 2456s 259s/step - loss: 0.5331 - accuracy: 0.9045 - val_loss: 1.0351 - val_accuracy: 0.7000\n",
      "Epoch 27/70\n",
      "10/10 [==============================] - 1718s 176s/step - loss: 0.4633 - accuracy: 0.9256 - val_loss: 1.0393 - val_accuracy: 0.6700\n",
      "Epoch 28/70\n",
      "10/10 [==============================] - 2294s 240s/step - loss: 0.5062 - accuracy: 0.9023 - val_loss: 1.0764 - val_accuracy: 0.7400\n",
      "Epoch 29/70\n",
      "10/10 [==============================] - 3369s 285s/step - loss: 0.4960 - accuracy: 0.9047 - val_loss: 1.0784 - val_accuracy: 0.7200\n",
      "Epoch 30/70\n",
      "10/10 [==============================] - 1249s 124s/step - loss: 0.5062 - accuracy: 0.9015 - val_loss: 0.9433 - val_accuracy: 0.7700\n",
      "Epoch 31/70\n",
      "10/10 [==============================] - 2984s 316s/step - loss: 0.4127 - accuracy: 0.9106 - val_loss: 1.1146 - val_accuracy: 0.6800\n",
      "Epoch 32/70\n",
      "10/10 [==============================] - 11554s 1172s/step - loss: 0.3830 - accuracy: 0.9444 - val_loss: 0.9391 - val_accuracy: 0.7500\n",
      "Epoch 33/70\n",
      "10/10 [==============================] - 2320s 242s/step - loss: 0.3761 - accuracy: 0.9436 - val_loss: 1.1654 - val_accuracy: 0.6600\n",
      "Epoch 34/70\n",
      "10/10 [==============================] - 2076s 215s/step - loss: 0.3925 - accuracy: 0.9514 - val_loss: 1.0021 - val_accuracy: 0.6900\n",
      "Epoch 35/70\n",
      "10/10 [==============================] - 1264s 126s/step - loss: 0.3434 - accuracy: 0.9613 - val_loss: 0.9509 - val_accuracy: 0.7400\n",
      "Epoch 36/70\n",
      "10/10 [==============================] - 1581s 161s/step - loss: 0.3540 - accuracy: 0.9558 - val_loss: 1.0313 - val_accuracy: 0.7300\n",
      "Epoch 37/70\n",
      "10/10 [==============================] - 1243s 123s/step - loss: 0.3497 - accuracy: 0.9641 - val_loss: 1.0081 - val_accuracy: 0.7200\n",
      "Epoch 38/70\n",
      "10/10 [==============================] - 1365s 138s/step - loss: 0.3213 - accuracy: 0.9557 - val_loss: 1.0416 - val_accuracy: 0.6500\n",
      "Epoch 39/70\n",
      "10/10 [==============================] - 3114s 319s/step - loss: 0.3412 - accuracy: 0.9643 - val_loss: 1.0064 - val_accuracy: 0.7100\n",
      "Epoch 40/70\n",
      "10/10 [==============================] - 2395s 251s/step - loss: 0.2999 - accuracy: 0.9718 - val_loss: 0.9560 - val_accuracy: 0.7300\n",
      "Epoch 41/70\n",
      "10/10 [==============================] - 1452s 145s/step - loss: 0.2922 - accuracy: 0.9703 - val_loss: 1.0067 - val_accuracy: 0.7200\n",
      "Epoch 42/70\n",
      "10/10 [==============================] - 1393s 137s/step - loss: 0.2726 - accuracy: 0.9802 - val_loss: 1.0006 - val_accuracy: 0.7200\n",
      "Epoch 43/70\n",
      "10/10 [==============================] - 7325s 722s/step - loss: 0.3310 - accuracy: 0.9723 - val_loss: 0.9089 - val_accuracy: 0.7600\n",
      "Epoch 44/70\n",
      "10/10 [==============================] - 17628s 1743s/step - loss: 0.2695 - accuracy: 0.9830 - val_loss: 0.9610 - val_accuracy: 0.7300\n",
      "Epoch 45/70\n",
      "10/10 [==============================] - 16285s 1622s/step - loss: 0.3239 - accuracy: 0.9613 - val_loss: 0.9850 - val_accuracy: 0.7000\n",
      "Epoch 46/70\n",
      "10/10 [==============================] - 3137s 180s/step - loss: 0.2493 - accuracy: 0.9857 - val_loss: 0.9291 - val_accuracy: 0.7600\n",
      "Epoch 47/70\n",
      "10/10 [==============================] - 2024s 157s/step - loss: 0.2650 - accuracy: 0.9832 - val_loss: 0.9996 - val_accuracy: 0.7000\n",
      "Epoch 48/70\n",
      "10/10 [==============================] - 1610s 165s/step - loss: 0.3219 - accuracy: 0.9709 - val_loss: 0.9569 - val_accuracy: 0.7200\n",
      "Epoch 49/70\n",
      "10/10 [==============================] - 1330s 132s/step - loss: 0.2636 - accuracy: 0.9836 - val_loss: 0.9319 - val_accuracy: 0.7300\n",
      "Epoch 50/70\n",
      "10/10 [==============================] - 2143s 222s/step - loss: 0.2275 - accuracy: 0.9894 - val_loss: 0.9493 - val_accuracy: 0.7500\n",
      "Epoch 51/70\n",
      "10/10 [==============================] - 1617s 166s/step - loss: 0.2404 - accuracy: 0.9882 - val_loss: 0.9769 - val_accuracy: 0.7100\n",
      "Epoch 52/70\n",
      "10/10 [==============================] - 1384s 137s/step - loss: 0.2195 - accuracy: 0.9858 - val_loss: 0.9488 - val_accuracy: 0.7200\n",
      "Epoch 53/70\n",
      "10/10 [==============================] - 2333s 241s/step - loss: 0.2274 - accuracy: 0.9915 - val_loss: 0.9739 - val_accuracy: 0.7300\n",
      "Epoch 54/70\n",
      "10/10 [==============================] - 2544s 269s/step - loss: 0.2214 - accuracy: 0.9880 - val_loss: 0.9301 - val_accuracy: 0.7200\n",
      "Epoch 55/70\n",
      "10/10 [==============================] - 1352s 134s/step - loss: 0.2278 - accuracy: 0.9954 - val_loss: 0.9970 - val_accuracy: 0.7500\n",
      "Epoch 56/70\n",
      "10/10 [==============================] - 1462s 146s/step - loss: 0.2165 - accuracy: 0.9869 - val_loss: 0.9273 - val_accuracy: 0.7500\n",
      "Epoch 57/70\n",
      "10/10 [==============================] - 1387s 138s/step - loss: 0.2136 - accuracy: 0.9953 - val_loss: 1.0968 - val_accuracy: 0.6800\n",
      "Epoch 58/70\n",
      "10/10 [==============================] - 1734s 176s/step - loss: 0.2285 - accuracy: 0.9868 - val_loss: 0.9515 - val_accuracy: 0.7400\n",
      "Epoch 59/70\n",
      "10/10 [==============================] - 2793s 291s/step - loss: 0.2105 - accuracy: 0.9944 - val_loss: 0.8986 - val_accuracy: 0.7700\n",
      "Epoch 60/70\n",
      "10/10 [==============================] - 2322s 243s/step - loss: 0.2128 - accuracy: 0.9919 - val_loss: 0.9093 - val_accuracy: 0.7500\n",
      "Epoch 61/70\n",
      "10/10 [==============================] - 1533s 153s/step - loss: 0.1931 - accuracy: 0.9930 - val_loss: 0.9337 - val_accuracy: 0.7600\n",
      "Epoch 62/70\n",
      "10/10 [==============================] - 1583s 158s/step - loss: 0.1781 - accuracy: 0.9963 - val_loss: 0.8980 - val_accuracy: 0.7800\n",
      "Epoch 63/70\n",
      "10/10 [==============================] - 1717s 172s/step - loss: 0.2032 - accuracy: 0.9941 - val_loss: 0.9277 - val_accuracy: 0.7600\n",
      "Epoch 64/70\n",
      "10/10 [==============================] - 1871s 187s/step - loss: 0.1887 - accuracy: 0.9981 - val_loss: 0.9176 - val_accuracy: 0.7500\n",
      "Epoch 65/70\n",
      "10/10 [==============================] - 1856s 182s/step - loss: 0.1997 - accuracy: 0.9980 - val_loss: 0.9167 - val_accuracy: 0.7500\n",
      "Epoch 66/70\n",
      "10/10 [==============================] - 1505s 150s/step - loss: 0.1927 - accuracy: 0.9923 - val_loss: 0.9353 - val_accuracy: 0.7300\n",
      "Epoch 67/70\n",
      "10/10 [==============================] - 2517s 265s/step - loss: 0.1691 - accuracy: 0.9957 - val_loss: 1.0022 - val_accuracy: 0.7200\n",
      "Epoch 68/70\n",
      "10/10 [==============================] - 3250s 343s/step - loss: 0.2114 - accuracy: 0.9960 - val_loss: 0.8853 - val_accuracy: 0.7700\n",
      "Epoch 69/70\n",
      "10/10 [==============================] - 8537s 640s/step - loss: 0.1725 - accuracy: 0.9985 - val_loss: 0.9083 - val_accuracy: 0.7500\n",
      "Epoch 70/70\n",
      "10/10 [==============================] - 8844s 765s/step - loss: 0.1735 - accuracy: 0.9978 - val_loss: 0.9332 - val_accuracy: 0.7500\n"
     ]
    }
   ],
   "source": [
    "hybrid = model.fit(X_train, \n",
    "                   y_train,\n",
    "                   epochs = 70,\n",
    "                   batch_size = 64,\n",
    "                   shuffle = True, \n",
    "                   validation_data = (X_test, y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "07a2a88f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten (Flatten)            (None, 784)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 392)               307720    \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 196)               77028     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 98)                19306     \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 49)                4851      \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 30)                1500      \n",
      "_________________________________________________________________\n",
      "keras_layer (KerasLayer)     (None, 4)                 128       \n",
      "=================================================================\n",
      "Total params: 410,533\n",
      "Trainable params: 410,533\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "49b07412",
   "metadata": {},
   "source": [
    "## 8. Loss and accuracy graphs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "cfd7c39c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# ===================================================================================\n",
    "#                                  Loss History Plot\n",
    "# ===================================================================================\n",
    "\n",
    "plt.title('model loss')\n",
    "plt.plot(hybrid.history['loss'], '-g')\n",
    "plt.ylabel('loss')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "cc6435a2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# ===================================================================================\n",
    "#                                Accuracy History Plot\n",
    "# ===================================================================================\n",
    "\n",
    "plt.title('model accuracy')\n",
    "plt.plot(hybrid.history['accuracy'], '-g')\n",
    "plt.ylabel('accuracy')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}