AlexNet-Baseline-wDropout.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Mini AlexNet Baseline w/Dropout Results\n",
    "\n",
    "### use Batchnorm and Dropout on all layers\n",
    "\n",
    "Expects:\n",
    "\n",
    "- weight.h5 files in ./weights \n",
    "- CIFAR10 data in ../data/cifar-10-batches-py/\n",
    "- Layer 8 becomes Layer 9 here\n",
    "\n",
    "and while the keras code needs py27, the notebooks can use py3.6 \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:48:41.680621Z",
     "start_time": "2018-09-02T22:48:38.326772Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/importlib/_bootstrap.py:219: RuntimeWarning: compiletime version 3.5 of module 'tensorflow.python.framework.fast_tensor_util' does not match runtime version 3.6\n",
      "  return f(*args, **kwds)\n",
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "importing Jupyter notebook from RMT_Util.ipynb\n",
      "3.6.5 |Anaconda, Inc.| (default, Apr 26 2018, 08:42:37) \n",
      "[GCC 4.2.1 Compatible Clang 4.0.1 (tags/RELEASE_401/final)]\n",
      "numpy version 1.14.3\n",
      "scipy version 1.1.0\n",
      "tensforflow version 1.10.1\n",
      "keras version 2.2.2\n",
      "sklearn version 0.19.1\n",
      "powerlaw version 1.4.3\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "import pickle, time\n",
    "from copy import deepcopy\n",
    "from shutil import copy\n",
    "from tqdm import tqdm_notebook as tqdm\n",
    "\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Flatten\n",
    "from keras.layers.convolutional import Conv2D, MaxPooling2D\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.callbacks import TensorBoard, EarlyStopping\n",
    "from keras.optimizers import SGD\n",
    "from keras.initializers import Constant\n",
    "import keras.backend as K\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import sklearn\n",
    "from sklearn.decomposition import TruncatedSVD\n",
    "from sklearn.random_projection import sparse_random_matrix\n",
    "\n",
    "import import_ipynb\n",
    "import RMT_Util"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "DataDIR = \"/Users/charleshmartin/work/DLGeneralization/data\"\n",
    "ModelDIR = \"/Users/charleshmartin/work-tmp/MiniAlexNet\"\n",
    "WeightsDIR = ModelDIR+\"/weights\"\n",
    "ResultsDIR = ModelDIR+\"/results\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:48:41.743119Z",
     "start_time": "2018-09-02T22:48:41.682478Z"
    }
   },
   "outputs": [],
   "source": [
    "DATA_DIR = DataDIR\n",
    "\n",
    "BASELINE_outfile = ResultsDIR+'/cifar10_alexnet.out'\n",
    "\n",
    "BASELINE_final_weightfile = WeightsDIR+'/alexnet.b16.final.h5'\n",
    "BASELINE_per_epoch_weightfile = WeightsDIR+\"/alexnet.b16.e{}.h5\"\n",
    "NUM_BASELINE_EPOCHS = 100\n",
    "\n",
    "\n",
    "DROPOUT_outfile = ResultsDIR+'/cifar10_alexnet.dp.out '\n",
    "\n",
    "DROPOUT_final_weightfile = WeightsDIR+'/alexnet.dp.b16.final.h5'\n",
    "DROPOUT_per_epoch_weightfile = WeightsDIR+\"/alexnet.dp.b16.e{}.h5\"\n",
    "NUM_DROPOUT_EPOCHS = 100\n",
    "\n",
    "DROPOUT_ensemble_weightfile = WeightsDIR+\"/alexnet.dp.b16.id{}.h5\"\n",
    "\n",
    "DROPOUT_init_weighfile = WeightsDIR+'/alexnet.dp.b16.e0.h5'\n",
    "\n",
    "NUM_ENSEMBLE_RUNS = 10\n",
    "N7, M7 = 4096, 384\n",
    "N8, M8 = 384, 192\n",
    "\n",
    "Q7 = float(N7)/float(M7)\n",
    "Q8 = float(N8)/float(M8)\n",
    "\n",
    "LAYER_FC1 = 7\n",
    "LAYER_FC2 = 9\n",
    "\n",
    "ENTROPY_FILE = \"img/mini-alexnet-dropout-entropy-per-epoch.png\"\n",
    "RANK_FILE = \"img/mini-alexnet-dropout-rank-per-epoch.png\"\n",
    "ACCURACY_FILE = \"img/mini-alexnet-dropout-accuracy-per-epoch.png\"\n",
    "\n",
    "\n",
    "TITLE_FC1 = r\"Mini AlexNet w/Dropout FC1: $\\rho_{NR}(\\lambda)$ \"\n",
    "TITLE_FC2 = r\"Mini AlexNet w/Dropout FC2: $\\rho_{NR}(\\lambda)$ \"\n",
    "\n",
    "\n",
    "TITLE_RAND_FC1 = r\"Mini-AlexNet w/Dropout FC1:\"+\"\\n\"+r\"$\\rho_{NR}(\\lambda)$ shuffled 100X\"\n",
    "TITLE_RAND_FC2 = r\"Mini-AlexNet w/Dropout FC2:\"+\"\\n\"+r\"$\\rho_{NR}(\\lambda)$ shuffled 100X\"\n",
    "\n",
    "XLABEL_FC1 = r\"Eigenvalues $\\lambda$ of $\\mathbf{X}=\\mathbf{W}_{FC1}^{T}\\mathbf{W}_{FC1}$\"\n",
    "XLABEL_FC2 = r\"Eigenvalues $\\lambda$ of $\\mathbf{X}=\\mathbf{W}_{FC2}^{T}\\mathbf{W}_{FC2}$\"\n",
    "\n",
    "XLABEL = r\"Eigenvalues $\\lambda$ of $\\mathbf{X}=\\mathbf{W}}^{T}\\mathbf{W}$\"\n",
    "YLABEL = r\"$\\rho_{NR}(\\lambda)$ shuffled $100X$\"\n",
    "\n",
    "FC1_FILENAME_1 = \"img/mini-alexnet-dropout-fc1-9spikes.png\"\n",
    "FC2_FILENAME_1 = \"img/mini-alexnet-dropout-fc2-9spikes.png\"\n",
    "\n",
    "FC1_FILENAME_2 = \"img/mini-alexnet-dropout-fc1-10spikes.png\"\n",
    "FC2_FILENAME_2 = \"img/mini-alexnet-dropout-fc2-10spikes.png\"\n",
    "\n",
    "RAND_FC1_FILENAME = \"img/mini-alexnet-dropout-fc1-rand.png\"\n",
    "RAND_FC2_FILENAME = \"img/mini-alexnet-dropout-fc2-rand.png\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:48:41.749255Z",
     "start_time": "2018-09-02T22:48:41.745216Z"
    }
   },
   "outputs": [],
   "source": [
    "matplotlib.rcParams.update({'font.size': 14})\n",
    "plt.rcParams['figure.figsize'] = [5,5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Check weights files are present"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:48:42.600999Z",
     "start_time": "2018-09-02T22:48:42.420483Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ls: WeightsDIR/*dp*: No such file or directory\r\n"
     ]
    }
   ],
   "source": [
    "!ls WeightsDIR/*dp* | head"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Baseline (first, for comparison)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:48:42.938818Z",
     "start_time": "2018-09-02T22:48:42.604576Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 28, 28, 96)        7296      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 10, 10, 96)        0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 10, 10, 96)        384       \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 10, 10, 256)       614656    \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 4, 4, 256)         0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 4, 4, 256)         1024      \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 4096)              0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 384)               1573248   \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 192)               73920     \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 10)                1930      \n",
      "=================================================================\n",
      "Total params: 2,272,458\n",
      "Trainable params: 2,271,754\n",
      "Non-trainable params: 704\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model = RMT_Util.Alexnet_Model()\n",
    "\n",
    "sess = tf.InteractiveSession()\n",
    "sess.run(tf.global_variables_initializer())\n",
    "\n",
    "model.load_weights(BASELINE_final_weightfile)\n",
    "print(model.summary())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py:1645: UserWarning: An interactive session is already active. This can cause out-of-memory errors in some cases. You must explicitly call `InteractiveSession.close()` to release resources held by the other session(s).\n",
      "  warnings.warn('An interactive session is already active. This can '\n"
     ]
    }
   ],
   "source": [
    "opt = SGD(lr=0.01, momentum=0.9)\n",
    "model.compile(loss='mse', optimizer=opt, metrics=['accuracy'])\n",
    "\n",
    "sess = tf.InteractiveSession()\n",
    "sess.run(tf.global_variables_initializer())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading test images...\n"
     ]
    }
   ],
   "source": [
    "test_images, test_labels = RMT_Util.read_test_images(datadir=DATA_DIR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "16bec4b7fd7141d384ebd3329f594f15",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 9s 939us/step\n",
      "10000/10000 [==============================] - 9s 940us/step\n",
      "10000/10000 [==============================] - 9s 917us/step\n",
      "10000/10000 [==============================] - 9s 915us/step\n",
      "10000/10000 [==============================] - 9s 936us/step\n",
      "10000/10000 [==============================] - 10s 986us/step\n",
      "10000/10000 [==============================] - 10s 1ms/step\n",
      "10000/10000 [==============================] - 10s 986us/step\n",
      "10000/10000 [==============================] - 10s 991us/step\n",
      "10000/10000 [==============================] - 10s 950us/step\n",
      "10000/10000 [==============================] - 9s 940us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 10s 978us/step\n",
      "10000/10000 [==============================] - 10s 954us/step\n",
      "10000/10000 [==============================] - 10s 970us/step\n",
      "10000/10000 [==============================] - 10s 979us/step\n",
      "10000/10000 [==============================] - 10s 993us/step\n",
      "10000/10000 [==============================] - 9s 933us/step\n",
      "10000/10000 [==============================] - 9s 918us/step\n",
      "10000/10000 [==============================] - 9s 904us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 917us/step\n",
      "10000/10000 [==============================] - 9s 930us/step\n",
      "10000/10000 [==============================] - 9s 919us/step\n",
      "10000/10000 [==============================] - 9s 910us/step\n",
      "10000/10000 [==============================] - 10s 963us/step\n",
      "10000/10000 [==============================] - 9s 938us/step\n",
      "10000/10000 [==============================] - 10s 993us/step\n",
      "10000/10000 [==============================] - 10s 961us/step\n",
      "10000/10000 [==============================] - 9s 923us/step\n",
      "10000/10000 [==============================] - 9s 904us/step\n",
      "10000/10000 [==============================] - 9s 911us/step\n",
      "10000/10000 [==============================] - 9s 904us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 910us/step\n",
      "10000/10000 [==============================] - 9s 899us/step\n",
      "10000/10000 [==============================] - 9s 904us/step\n",
      "10000/10000 [==============================] - 9s 898us/step\n",
      "10000/10000 [==============================] - 9s 906us/step\n",
      "10000/10000 [==============================] - 9s 904us/step\n",
      "10000/10000 [==============================] - 9s 898us/step\n",
      "10000/10000 [==============================] - 9s 916us/step\n",
      "10000/10000 [==============================] - 9s 917us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 902us/step\n",
      "10000/10000 [==============================] - 9s 896us/step\n",
      "10000/10000 [==============================] - 9s 899us/step\n",
      "10000/10000 [==============================] - 9s 906us/step\n",
      "10000/10000 [==============================] - 9s 920us/step\n",
      "10000/10000 [==============================] - 9s 928us/step\n",
      "10000/10000 [==============================] - 9s 936us/step\n",
      "10000/10000 [==============================] - 9s 918us/step\n",
      "10000/10000 [==============================] - 9s 905us/step\n",
      "10000/10000 [==============================] - 9s 910us/step\n",
      "10000/10000 [==============================] - 9s 928us/step\n",
      "10000/10000 [==============================] - 9s 918us/step\n",
      "10000/10000 [==============================] - 9s 909us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 903us/step\n",
      "10000/10000 [==============================] - 9s 897us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 899us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 939us/step\n",
      "10000/10000 [==============================] - 10s 952us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 932us/step\n",
      "10000/10000 [==============================] - 9s 898us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 894us/step\n",
      "10000/10000 [==============================] - 9s 941us/step\n",
      "10000/10000 [==============================] - 10s 966us/step\n",
      "10000/10000 [==============================] - 10s 955us/step\n",
      "10000/10000 [==============================] - 10s 980us/step\n",
      "10000/10000 [==============================] - 9s 929us/step\n",
      "10000/10000 [==============================] - 9s 902us/step\n",
      "10000/10000 [==============================] - 9s 899us/step\n",
      "10000/10000 [==============================] - 9s 901us/step\n",
      "10000/10000 [==============================] - 9s 898us/step\n",
      "10000/10000 [==============================] - 9s 908us/step\n",
      "10000/10000 [==============================] - 9s 920us/step\n",
      "10000/10000 [==============================] - 9s 912us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 896us/step\n",
      "10000/10000 [==============================] - 9s 898us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 900us/step\n",
      "10000/10000 [==============================] - 9s 908us/step\n",
      "10000/10000 [==============================] - 9s 902us/step\n",
      "10000/10000 [==============================] - 9s 906us/step\n",
      "10000/10000 [==============================] - 9s 899us/step\n",
      "10000/10000 [==============================] - 9s 902us/step\n",
      "10000/10000 [==============================] - 9s 902us/step\n",
      "10000/10000 [==============================] - 9s 899us/step\n",
      "10000/10000 [==============================] - 9s 896us/step\n",
      "\n"
     ]
    }
   ],
   "source": [
    "bsln_train_accs = RMT_Util.read_training_accuracies(BASELINE_outfile)\n",
    "assert(len(bsln_train_accs)==NUM_BASELINE_EPOCHS)\n",
    "\n",
    "bsln_test_accs = []\n",
    "for epoch in tqdm(range(NUM_BASELINE_EPOCHS)):\n",
    "    modelfile = BASELINE_per_epoch_weightfile.format(epoch)\n",
    "    model.load_weights(modelfile)\n",
    "    acc =  model.evaluate(np.array(test_images),np.array(test_labels), batch_size=256)\n",
    "    bsln_test_accs.append(acc[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-04-11T00:20:35.143766Z",
     "start_time": "2018-04-11T00:20:35.141086Z"
    }
   },
   "source": [
    "## Entropy and Accuracies per Epoch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:53:44.605609Z",
     "start_time": "2018-09-02T22:48:42.941336Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2420b4d0b94947039de3705cdcf8b8b2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "fc1_bsln_entropies, fc2_bsln_entropies = [], []\n",
    "fc1_bsln_ranks, fc2_bsln_ranks = [], []\n",
    "\n",
    "for epoch in tqdm(range(NUM_BASELINE_EPOCHS)):\n",
    "    model.load_weights(BASELINE_per_epoch_weightfile.format(epoch))   \n",
    "    fc1_bsln_entropies.append(RMT_Util.layer_entropy(model,LAYER_FC1))\n",
    "    fc2_bsln_entropies.append(RMT_Util.layer_entropy(model,LAYER_FC2-1)) # note\n",
    "    \n",
    "    fc1_bsln_ranks.append(RMT_Util.layer_soft_rank(model,LAYER_FC1))\n",
    "    fc2_bsln_ranks.append(RMT_Util.layer_soft_rank(model,LAYER_FC2-1)) # note"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:53:44.624543Z",
     "start_time": "2018-09-02T22:53:44.612704Z"
    }
   },
   "outputs": [],
   "source": [
    "assert(len(fc1_bsln_entropies)==NUM_BASELINE_EPOCHS)\n",
    "assert(len(fc2_bsln_entropies)==NUM_BASELINE_EPOCHS)\n",
    "\n",
    "assert(len(fc1_bsln_ranks)==NUM_BASELINE_EPOCHS)\n",
    "assert(len(fc2_bsln_ranks)==NUM_BASELINE_EPOCHS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Baseline  w/Dropout \n",
    "\n",
    "### (Batchnorm on FC1 and FC2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:53:46.378957Z",
     "start_time": "2018-09-02T22:53:44.627751Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py:1645: UserWarning: An interactive session is already active. This can cause out-of-memory errors in some cases. You must explicitly call `InteractiveSession.close()` to release resources held by the other session(s).\n",
      "  warnings.warn('An interactive session is already active. This can '\n"
     ]
    }
   ],
   "source": [
    "model = RMT_Util.Alexnet_Model(batchnorm=True)\n",
    "sess = tf.InteractiveSession()\n",
    "sess.run(tf.global_variables_initializer())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:58:41.312483Z",
     "start_time": "2018-09-02T22:53:46.381726Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9f1134199fc245938045337b7aba9e45",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "fc1_entropies, fc2_entropies = [], []\n",
    "fc1_ranks, fc2_ranks = [], []\n",
    "\n",
    "for epoch in tqdm(range(NUM_DROPOUT_EPOCHS)):\n",
    "    model.load_weights(DROPOUT_per_epoch_weightfile.format(epoch))   \n",
    "    fc1_entropies.append(RMT_Util.layer_entropy(model, LAYER_FC1))\n",
    "    fc2_entropies.append(RMT_Util.layer_entropy(model, LAYER_FC2))\n",
    "    \n",
    "    fc1_ranks.append(RMT_Util.layer_soft_rank(model, LAYER_FC1))\n",
    "    fc2_ranks.append(RMT_Util.layer_soft_rank(model, LAYER_FC2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:58:41.324487Z",
     "start_time": "2018-09-02T22:58:41.316494Z"
    }
   },
   "outputs": [],
   "source": [
    "assert(len(fc1_entropies)==NUM_DROPOUT_EPOCHS)\n",
    "assert(len(fc2_entropies)==NUM_DROPOUT_EPOCHS)\n",
    "\n",
    "assert(len(fc1_ranks)==NUM_DROPOUT_EPOCHS)\n",
    "assert(len(fc2_ranks)==NUM_DROPOUT_EPOCHS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:58:41.871741Z",
     "start_time": "2018-09-02T22:58:41.328486Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "saved file img/mini-alexnet-dropout-entropy-per-epoch.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(range(0,len(fc1_entropies)), fc1_entropies, marker='.', color='cyan', label=\"FC1 dropout\")\n",
    "plt.scatter(range(0,len(fc2_entropies)), fc2_entropies, marker='.', color='lime', label=\"FC2 dropout\")\n",
    "plt.scatter(range(0,len(fc1_bsln_entropies)), fc1_bsln_entropies, marker='.', color='blue', label=\"FC1 baseline\")\n",
    "plt.scatter(range(0,len(fc2_bsln_entropies)), fc2_bsln_entropies, marker='.', color='green', label=\"FC2 baseline\")\n",
    "\n",
    "#lt.ylim([0.905,0.920])\n",
    "plt.title('Mini-AlexNet w/Dropout:'+'\\n'+'Entropy vs Epoch')\n",
    "plt.xlabel(\"Epoch (e)\")\n",
    "plt.ylabel(r\"Matrix Entropy $\\mathcal{S}(\\mathbf{W}^{e}_{l})$\")\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(ENTROPY_FILE)\n",
    "print(\"saved file {}\".format(ENTROPY_FILE))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:58:42.412691Z",
     "start_time": "2018-09-02T22:58:41.874144Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "saved file img/mini-alexnet-dropout-rank-per-epoch.png\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVcAAAFYCAYAAADnZ7/NAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzsnXl8VOX1h5+TQEIQBGQHQVAQQVCruMQFoiBCqXtbt4po6wKi4s+VWhSLlYraqq1UwIqiXbRibQWlCBJBDShYrApuLILsirJLIDm/P947yZ3LTDKTzCQzk/PwuZ+Z+973vvfcYfKdc8997zmiqhiGYRiJJau2DTAMw8hETFwNwzCSgImrYRhGEjBxNQzDSAImroZhGEnAxNUwDCMJmLimOSKySkSeruK+Y0Qk4XPxRKSTiKiIDE302IaRLpi4pgAiMtQTIxWR06P0edfb/klN2xfBlimeLX+uZTtCn9kdEbZd7G0rqMK4B3s/PMckxNDYjvmliNzrvS/0nVupiGwTkU9F5FkRObOmbKpJauMzTzYmrqnF98BlwUYR6Qoc720P0g24uorHuw/Ii2cHEWkAXACsAi701mubW0XkgASOdzBwD1Ajf+gi0gvoCMzwNa8HLgeGALcB/wZOBmaJyPMiUr8mbKtBavQzrwlMXFOLGcCPRSQ30P4zYCOwKLiDqu5R1b1VOZiq7lPVSIJdEecCjXF/9E2As6ty7ASyBGgBjKhlO6rDYGAT8J6vbZuqPuctE1X1NuBwYALwU+A3FQ0ojlT44auzmLimFn/DCdePAu2XAn8HSoI7BGOuIlLgXU5eIiL/523/XkSKROQHgX2rEnP9GTBfVecDC731mBCRtiIySUTWicgeEflcRG4XEfH1udezf2Bg3996l8jBsMlCYBYxeq8i0lVE/i4imz0bPhSRK33bC4AibzUU/lARGRNlPBGRr0VkQqB9jbffob62UKjiiMAwPwJe00qeRVfVEuBGYBkwQkSa+MZWEXlCRH4iIv8D9gAXe9uyROQOEfnMO+e1IvIH//5ev0IR+UREjhaReSKyyzuPSGGXPO//ZJU35koRuS/oGET77LxjFXrvC6jgMxeR+iJyhIi0rejzSTVMXFOLtcCb+EIDInIi0AX4S5xj3QJcAfwBGIPzel6uzuWkiLQAzsIJPcBfgUEi0jyGfVsBC3BCMgknEguBB4Df+7qOBRYDfxaRZt6++cCtwGOqOjfC8PcQg/cqIt29Yx4L/A4YCXwJPCUiI71uy3CfF56dl3vLS5HG9ATxLaCv7zidcZe5pUAfX/c+wGZV/cTX9yDgJGB6Rbb7jleC+9zzgFMDm08D/uTZeiMQOs4E4Lfeuf0fLsQwHHg9wvehCfAf4GPgdm+M34rIL302i3eMO3Df15tx4ngX5d+NeKjsM2/v9RlXhbFrD1W1pZYXYCiguD+yn+O8jmbetj8An3nvC4FPAvuuAp72rRd4Y30ONPC1n+e1D/a1jcHThxjtHAHsBVp6622AfcCwQL9O3rGG+tom4UIbrQN9x+M88k6+th7AbtwPSkPvXD4B8gL7KvCE9/4/wGbgAG/9Ym97ga//LGAp0DAwzvPANt++JwXtr+Rz+T+ckLbw1ocAW4GZwJ99/T4CpgX2vRQoBg70te33/xzYJ/R/eWPgsygFjgn07eltezbQfpPX/ovAcRUY5WvLAuYCu4AmXtuPvH5jA2P+3mvvH7BrTIRzKAQKfetRP3Pf9+np4LZUXsxzTT1exH2Rfiwi9XDxtXi9VoApGh5PfdN7PTRS5xj5GfCGqm4GUNUNuD+8CkMDnqfzY5x3ViIiLUILThSz8Hl+qroU5wVdCryB++Maoqq7KzhMhd6r5wX3B14AGgZseA0Xjuld8elHZT4gOM8RnIf6Nk5A+njHb4770Zgf2PdHwFuqui2O4+3wXhsH2t9R1SURxgd4MND+BO4HZXCgvRR4PLSiqqH1PCAUkgmJ68OBfcd7r8Exq4WqrlJVUdWhiRw32Zi4phiquhV3Y+sy4EygFe4yMF6+DIz7rff2oGg7iEgjEWnjW1r6tnUBTgQKxc1j7SQinXDierI/thiBlkAz4Cqcd+lfZnt9WgX2eQR3g+dE4EFVfbeik1XVBVQce+2KE8B7ItgwJYoNsfI+TvBCIYDTgHne0kVE2nht4rUBICLZuDBLTCEBH4281+2B9uUR+nbCCeGn/kZV3QOs8Lb72RhB6D/zXg/xjblBVb8LjLke+C7CmHWSerVtgBGRv+A8WIB3VfXzKoyx380vD4nSDi6ueY9v/UvK/1Au917v95Ygl+HipZEI/Yj/DXgqSp/gOR4CdPfe94yyT5B7cLG/EQR+XHw2/B54Ncr+H8Z4nDBUtUREioA+ItIaF9+eh5vdsRsnuifhPMUPfLvm437sZhAfoc/ji0B7RZ59JAQnvH4i3VSr6DtT1b7ZRP+OZgQmrqnJDFzMri8uNlZTTMXdnAnh/2O9DCcYv2d/bsKFBqKJ62acsNRT1dlR+pThhRGm4GKR9wD3isiVqjqlov1UdYGIzML9SNwZ2LzCe90Xgw1VeWptHi6GPRj3uS1S1WIRWYjzWvNxl+1+QfkRsFxVPw0OFg3P270UFwN9O4ZdVuEErxvwP984OUBnXNjFTxsROTDgvXb1XkM/WKuAASLS1O+9eh56E297iG+BphHs6kS4p51xWfstLJCCeJdsw4F7cd5eTR13harO9i1vA4jIycBhwFRVfTm44ET5cBE5Psq4JThP/HwROTa4XUSaBO5aj8T9sAxX1V/jbgw9IiIdYziNUOz1+oANm3BCco2IdIhgQ0vf6k7vtVkMxwsxD+eN3Q4sUNVir30+MAg3OX5eYJ/BxBES8IT1MZxH/5gXQqqMkFf8f4H2a4ED2d9rzsL32YlIFu67+D0uhoxns+BmCfi5LXBMcN51QeA8zsXNpvAT9TNP16lY5rmmKKpaY6IaAz/D3eiIJgQzvO0/I3wivJ87cYL5trjHZj/E/XH3BC7ETTfbICLdcBPkX1DV5719f4G70/6UiJyp3i3kSPi81wERNg/DeXv/E5HJuFBEC9zUrLM8e/DatwHDRGQHLrb5kap+FO24wLu4WR7dCJ+ONA8Y7b0vu5nl/VD0ZH/RC3GgiIRuFDbEfT4X4H7k/u4bs0JU9UMRmQhc681rfd077rW4/6tnArtsAG7y7PsINzPhdOBun5c6A/eDd7fX7z3c02OXAf8KXBlMBJ4UkZdxNw6PwHnewfhwRZ95aCrWM7iZNelBbU9XsCV8KlYl/QqJfSrWxRH2D5sWQwxTsYD6wNdAUSX9inBTreoRYSqW16c5LqywEnfJvwkXhrgVyMF5fgtxj342D+x7mTfmiMD5PBHBltC0nrCpWN62jsCTuDnFxcA63E21awL9zsb9ABQHP7cKPoP5Xt8zfG0H4Kav7QZyfe3DcQKSE+X/WX3LdtxNpWeBM6McO+Jn4W3Lws1J/dx3zn/Em1oV/H4BR+N+FHYDXwGjAAn0zcPNUV7tjbkK9zh1boRj34cT7d3euMcSmIpV0WdOmk7FEs94wzBqEBGZAexR1Qtq25YQ3hNTbVQ1+ASZUQUsLGAYtcObxHZDykhTTFwNoxZQ1fGV9zLSGZstYBiGkQQs5moYhpEEzHM1DMNIAiauhmEYScDE1aiUUCLmGPqFEnUX1IBZNYKU1zc7qbZtSUdE5GkRibfaRUZg4pqmiEgvEXlByisNrBORN4NZ30VkhKRxFVYJL9anIrJbRP4nIiO9RzPrNIHPJri8WPkIRrKwqVhpiIicgntOfj3wNO4pmna4J5N+RXlWd3AZojZ4/dKV9bhn9sE9rnop7kmvVsAvo+1Uh3iD8rSJfoKZwYwaxMQ1PfklLtFFb1X92r9BRNrVjklJZZuqPhda8UIUn+DqSI3W8ExTdZHP/Z+PkRrU+cuqNOUw4OOgsAKo6rrQexFZhUsk0td3qbjK25YjIr8WkUUi8q13ub1QRM6JdlARuUhElnphiP+JSEwZ56WSooDxoq7Cwnu4TPxlCa7jOScpL+g3SET+653T5yJySQzn08gLV3wdKcuX16e1iOwTkf3SMEbaJiLDvc9lp4hsEZHFInJdjB9JpYRinyJyiIjMEJEdIrJRRMYHMpLFXNDQ69tbRF7xbN4lIh+JyKgo5/y8iGzz/m8mSYZXpzVxTU9WAceIyNGV9BuJu6T+hPKib6FCfAcC11FeWO5XuCQtL4vIWRHGOgVX/O4Fr3+u17dPhL5lSGxFAatCJ1wyD382/HjP6STc5fS/cMljdgLPeTZHO5+muEQvR+CSwrwfqZ+qbsQlJ7kowuaf4JLU/N0b8+e4Uiqf4rJk3Y3LshUsQBiNBuIrW+NbguKVhctmtQ2XyGUeLk3g44F+MRU0FJF+uMQ7x+ISwdyMqwYR/DHL8tqLceGdl4GrvfPMXGo7c4wt8S/AGbgs7iU44XoIlxu0QYS+nxDIPuS1Z7N/BqMcXAG/1wPtoexMJ/vamgNbcAmgQ20FVLEoYAXnWkh5asAWOE98vHecl6t5TnuB7r621ri0gQ/62oZSXjyyBa6kyxrg8Bj+n37u7XtcoP0t4EPf+j9x6fWq8l3QChZ/BrGnvbaJgf2fwaWL7Oatx1TQECeYy3EZtloG+kqE494X6PMyrhJurf89JWsxzzUNUdU3cNntX8EVvbsFl2t1Y6yX26paoi4pd+hy+iCc5/cmcFyEXRap6ju+/b/B1fbKF68EdhBJXFHALpTXu/oE521NA8LOtQrnNFdVl/n23+iNH6keWFtvnKZAH1X9LEKfIC/hvLUy71VEDsblPvXnfN0GHCwiJ8QwZiSm4+qtBZeXI/R9LLD+B1zi60HeeqwFDY/FfU6PqFewMoR66hlgQmD9TaCFiASLLGYMdkMrTfGE7jxx2emPwv1R3IK73P7SE+AKEZFf4C7luhNe+yjSH0ekOl4hgemIK+cRxF8U8J4I2yG2ooBrcMUNs3BC+0vc7IjiYMc4zynS3fRviVzEcar32kNV18RgM6r6rbjE3ReJyB2e6Fzk2eUX1weAfsBCEVmBS2j9Qiz/hx5rNYbyObjPIFh3K1LxwYgFDT3bOnlNh3mvsdQd26u+ewEe/oKZwUKLGYF5rmmO5639V1XH4jLVQyWlrgG8GzeTcX9sQ3Gey5k4bzRSkbmqFK7zFwWM5Fmdyf6lTyKxS13ZmVmqOsGz9XgChRKrcE7xFHF8EVd19dYY7PXzN9yPT763fjHuKqAsE7+6UuLdcLHYN3A/lHMkhgc3ahiJ8D6W5CSlMY6ZUZjnmlmEyk/7p2NF+/JfhCvad57/Mq6CsMLhEdpChetWR9knnqKAMaOudMkzuJIgv1PVkAca7znFw0TgY+BBEdmpqrHOr/0XrpjgxSKyERcG2U+gVXUnTsBfFJF6uFjotSIyVlXXJsB+cELWBXceIUL/r/7ig7EUNAx5wL1wN8mMAOa5piEicoZEfjrph97rJ762nUQutBfy2so8BxE5FDg/ymF7i0i+r29z3GT+BaoaKSSAxlcUMF4exDkHt/na4j2nuFDVh3BFI0eJSEzi6onmdJxXeinux+4Ffx/vs/Tvs49yYYunSGIs3BhYv8F7DQlkrAUN38f9kI0M2i8iGeuNxoN5runJY0AjEfknbrpMFu4Gw+XAN8Ajvr6LcOJ2Dy6+tkNVX8FNr7kA+LeI/BtXBG641yfSFK+PgOki8gdcjOwa3B/bfnMaA8RaFDAuVPVTEZkOXCUiv/aEPN5zqspxx4jIAcBvRGSHqgZvEEXi78BPcT8Eb0eI2c4SkVA9sQ047/IGXDxzaQzjd5XyYoZ+vlNVf1HJvUAfEfmrd6zTgR8DT6nqJ975xVTQUFVLvXm4M4AlIvIUri5ZF9y0vVNisDuzqe3pCrbEvwADcbHFpbg7uHtwRf+eBDoH+rbBic5WnNe0yrftNpz38T1OPH9GhKKF3n5P4C67l3nH+wg4O9CvgGoUBYxyroUEijL6tvXxjjeuqucU5XiFvvWhRCgeibv7XQpcFcM55OLm4ypwfYTt13jH3ez7v3wMaBXD2BVNxfrI1+9p7zM5BCeIO7zjPUSgSCIxFjT0+uZTPnd2J+4H4Y7gcSPsF/pcO9X231OyFkuWbRh1ABF5GlcROKOfikolLOZqGIaRBExcDcMwkoCJq2EYRhKwmKthGEYSMM/VMAwjCdg8V6BFixbaqVOn2jbDMIw0YPHixV+raqUPwJi4Ap06dWLRokW1bYZhGGmAiMRUPsfCAoZhGEnAxNUwDCMJmLgahmEkARNXwzCMJGDiahiGkQRstoBRJygtLeWrr75i586dtW2KkcLUr1+fVq1aceCBVcqEGYaJq1En+PrrrxERunXrRlaWXbAZ+6Oq7N69m7VrXeGH6gqsfcuMOsF3331H69atTViNqIgIDRs2pH379mzatKna49k3rQoUFcG4ce7VSA9KSkqoX79+bZthpAF5eXns3bu32uNYWCBOioqgXz8oLoacHJgzB/LzK9/PqH2stJMRC4n6npjnGieFhU5YS0rca2FhbVtkGEYqYuIaJwUFzmPNznavBQW1bZFh1AwjRoygwL7wMWPiGif5+S4UMHashQSM5DJ06FBEZL9lyZIlZX02btzITTfdxGGHHUZubi7t27dn0KBBvPrqq2V9Jk2axOmnn07Tpk0REVatWlULZ1PzrFq1ChGptaRMFnOtAvn5JqpGzdC/f3+effbZsLYWLVoATjxOOeUUGjduzLhx4zj66KMpLS1lzpw5XHfddaxevRqAXbt2MWDAAM4991xuvvnmpNlaXFxMTk5O0sZPN8xzNYwUJjc3lzZt2oQt9eo5n2j48OGoKosWLeKnP/0p3bp1o3v37owYMYIPPvigbIyRI0cyatQoTj311JiPW1JSwq233kqzZs1o1qwZI0eOpKSkJKxPQUEBw4YN49Zbb6Vly5accsopAKxevZrzzz+fxo0b07hxYy644AK++uqrsv3GjBlDz549efLJJ+nYsSN5eXmcd955fP3112V9SktLGTt2LB06dCA3N5devXrxr3/9q2x7NK9URHjxxRcB6Ny5MwDHH388IlLjIQ0TV8NIQ7Zs2cLMmTMZMWIEjRo12m97s2bNqjX+ww8/zOTJk5k4cSJFRUWUlJTwl7/8Zb9+zz33HKrK/PnzmTp1KqrKeeedx8aNG3njjTeYO3cu69at47zzzsNfUmrVqlU899xz/Otf/2L27Nl8/vnnXHXVVWXbH330UR588EEeeOABPvzwQ84//3wuuOCCsJBIZbz77rsAzJw5k/Xr1/PSSy9V4xOpAqpa55fjjjtOjcxm6dKlCRnnHVW933tNNldccYVmZ2frAQccULYMHDhQVVUXLlyogL700ksxj/fee+8poCtXrqy0b9u2bfW+++4rWy8pKdGuXbtq3759y9r69u2rvXr1Cttv1qxZmpWVFXaM5cuXq4jo66+/rqqq99xzj2ZlZemXX35Z1mf+/PkK6Geffaaqqu3atdN77703bOy+ffvqZZddpqqqK1euVEDfe++9sD6A/uMf/6iwTyxU9H0BFmkMumIxV8OIkSKgH1AM5ABzgGSH3vv06cOkSZPK1vPy8gDCvMBEs3XrVtavX0++78ZCVlYWJ554ImvWrAnre9xxx4WtL1u2jHbt2uEvm3TooYfSrl07li5dSv/+/QFo3749HTt2LOtz4oknkpWVxbJly2jdujXr1q0rCzOEOPXUU8Nu1KU6FhYwjBgpxAlrifdaWAPHbNiwIV26dClb2rdvD0DXrl0REZYtW1YDVkTngAMOCFtX1aiT8OOdnB+pf6gt9Biz/0cmEU9VJRITV8OIkQKcx5rtvRbUoi0HHXQQZ511Fn/84x/ZsWPHftu/++67Ko/dpEkT2rZty4IFC8raVLUshlkRPXr0YO3atWHTvVasWMG6devo0aNHWdvatWvDvOB3332X0tJSunfvzoEHHki7du146623wsZ+6623ysZo2dLVB1y/fn3Z9mA8NjRzIXgjrqawsIBhxEg+LhRQiBPW2p6NN2HCBE4++WR69+7N2LFjOeqoo1BV5s6dy7hx48qmYm3YsIENGzbw2WefAbB06VK+++47OnbsyEEHHRRx7Jtuuolx48Zx+OGH06tXLyZMmMD69etp27ZthTb179+fo48+mssuu4zHHnsMVeWGG27g2GOP5Ywzzijrl5eXxxVXXMHvfvc7du/ezXXXXcfgwYPp2rUrALfddht33303Xbt25bjjjuO5555j/vz5LF68uGz/k046iQceeIDDDjuMrVu3MmrUqDBbWrVqRV5eHv/5z3/o1KkTDRo0oEmTJlX7sKtCLIHZTF/shlbmk6gbWjXJFVdcoYMHD66wz7p163TEiBHauXNnzcnJ0bZt2+rAgQP11VdfLetzzz33KLDfMmXKlKjj7t27V0eOHKlNmjTRJk2a6IgRI/S6667b74bW9ddfv9++X375pZ577rnaqFEjbdSokZ533nm6Zs2aMHuOPPJInThxoh588MHaoEEDPeecc3TTpk1lfUpKSvTXv/61HnzwwVq/fn3t2bOn/vOf/ww7ztKlS/Xkk0/WvLw87dmzp86bNy/shpaq6uTJk7VDhw6alZUVZntlJOKGlmgSA+PpQu/evdVKa2c2y5Yto3v37rVthoGb5/riiy/y0Ucf1bYpUano+yIii1W1d2VjWMzVMAwjCZi4GoZhJAET12piibMNIz7GjBmT0iGBRGGzBaqBJc42DCMa5rlWA0ucbRhGNExcq4ElzjYMIxoWFqgGocTZhYVOWC0kYBhGCBPXamKJsw3DiISFBQzDMJJArYuriIwSkfdEZJuIbBaRV0SkZ6CPiMgYEVknIrtFpFBEjgz0aSYiz4rIVm95VkSa1uzZGEbmYgUK46PWxRWXA2MCcDJwBrAPmC0i/owStwO3ADcAxwObgNdFpLGvz1+BY4FBwEDvfXjxIcNIIxJRoHDLli3ccMMNHHHEEeTl5dGhQweGDRvGN998U1unVWPU+QKFqnqWf11ELge2AqcAr4hL4DgS+K2qTvP6XIET2EuBiSLSHSeop6rqO16fa4H5ItJNVT+tsRMyjARS3QKF69atY+3atYwfP74sHeDw4cO55JJLmDVrVkJttQKFAWLJ7lKTC9AWl7XnVG/9UG/9+EC/GcAz3vurgO3gEtF4bQLsAK6s7JiWFSvzycSsWIMGDdK2bdvq9u3b99u2ZcuWqPvNmDFDRUS3bt0atc++ffv0lltu0aZNm2rTpk31pptuipgV67rrrtNbbrlFW7Roob1791ZVlxXrvPPOK8uKdf7550fMihXKWNWgQQM999xzdfPmzWV9/FmxcnJytGfPnvryyy+XbY+lzAuBLGA1nRUrFcICQR4FluCqagC08V43Bvpt9G1rA2z2ThwA7/0mX58wROQaEVkkIos2b96cKNsNo0aoToHCbdu2kZubS8OGDaP2sQKFCSAWBa6pBfgdsA441Nd2Mu6Xp0Og7xRgpvf+l8DyCOOtBO6s7LjmuWY+iStQ+I7er/frOzVQojDRBQpVVb/99lvt0qWL3nDDDRX2swKFGVSgUER+D1wMnK6qK3ybNnivbQB/dbRWlHuzG4BWIiLeyePFaluyv8drGFWiiCL60Y9iiskhhznMIT/J9QgSWaBw586dnH322bRv357x48dH7WcFChNDSoQFRORR3M2pM1T1k8DmlTjxPNPXvwFwGvCO11QENCK88kY+cICvj2FUi0IKKaaYEkooppjCGihRmKgChTt27GDQoEEATJ8+nQYNGiTEPitQGJ1aF1cReRy4ErgE+FZE2nhLIyiLnT4C3CkiF3hzYJ/G3az6q9dnGTATN3PgJBHJByYC09VmChgJooACcsghm2xyyKGgFksUxlOgcPv27QwcOJCSkhJeffXViDFaP1agMDGkQlhguPc6J9B+LzDGez8eyAMeB5oBC4EBqrrd1/8y4DEgNL/k38CIJNhLEalTpM6oOfLJZw5zKKSQAgqSHhKojFgKFG7fvp0BAwawbds2Xn75ZXbu3MnOnTsBJ9DRpk5ZgcIEEEtgNtOXeG5ovaOqeaqa7b2Gbmu8847q/fe7VyP1yMSpWKqVFyicO3duxOKEgM6dOzfquFag0AoUJoR4ChSOA0YDJbj69WOBAkuanfJYgcLUwQoUGhEpAHJwwprjrVvSbMMwgqRCzDWtyMcFhwvxxVwLnMca8lwtt4VhGBYWIL6wQDSKiixpdipjYQEjHhIRFjDPNUFY0mzDMPxYzNUwDCMJmLgahmEkARNXwzCMJGDiahiGkQRMXA3DMJKAiathGPsxdOhQfvSjH9W2GYgIL774YtT1VMbE1TBSlEQUKASYNGkSp59+Ok2bNkVEwjJWpRvr16/n7LPPrm0zYiKuea4ikgu0w2Wo2qyqVh8lAvZAgZEoqlugEGDXrl0MGDCAc889l5tvvrnGzyGRtGkTsWpTSlKp5yoijUVkmIjMw1Vl/QL4CNggImtEZLKIHJ9sQ9OFIi+Jy+jR7rWoqPJ9DCMaubm5tGnTJmypV8/5RMOHD0dVWbRoET/96U/p1q0b3bt3Z8SIEXzwwQdlY4wcOZJRo0Zx6qmnxn38++67j9atW9OoUSOuvPJKdu/eXbZt5syZnHbaaTRr1qwsv2wwefevf/1rDjnkkLLzGDJkSNk2VWX8+PEcdthh5OXl0atXL5577rkK7fGHBUKls6dNm8aZZ55Jw4YN6dGjB6+//nrYPkuXLmXw4ME0btyYVq1acckll7Bhw4ZIwyeUCsVVRG4GVuGqq74OnAscAxyOe6z+Hpz3+7qIzBSRrkm1Ng2wJC5GTVCdAoWx8uabb/LBBx8wZ84cpk2bxqxZs7jjjjvKtu/cuZORI0fy7rvvUlhYSJMmTTj77LMpLi4GYNq0aTz00ENMmDCBzz//nOnTp3PCCSeU7f+rX/2KP//5zzz++OMsXbqUUaNGce211zJjxoy47Lzrrru48cYb+eCDDzj++OO5+OKLyxKIr1+/nj59+tCzZ0/effddZs+ezY4dOzjnnHMoLS2t9mdUIRXlIwT+AfSsLG8hkItLev2xxmkgAAAgAElEQVSLWPIcptqSyAKF77yjmpenmp3tXi2/a2qQsAKFNZi3N9EFCt977z0FwooHVnTsJk2ahJXtfvbZZzUnJ0d37NgRcZ8dO3ZoVlaWzp8/X1VVH374YT388MO1uLg4Yt8GDRrovHnzwtpvuukmHTRoUNk6gfysRChA+MQTT5Rt/+qrrxQos2H06NF6xhlnhB1jy5YtCujChQujnn/SCxSq6k9iFOg9wISqyXtmkZ/v8rlazDXzKKqFvL2JLFAYL0cddVSYV5yfn09xcTHLly/nqKOOYvny5YwePZqFCxeyefNmSktLKS0tLYv1/uQnP+HRRx+lc+fOnHXWWQwcOJBzzjmH3Nxcli5dyvfff8/AgQPDamXt3bs3rLhhrHaGaNeuHQCbNm0CYPHixcybNy+id798+fIwTzrRVCiuItIKVyqlWFXvS5oVaUykki+WxCUziRTySfb/c6hAYRB/gcLzzz8/uUZEIVRJduLEibRv35569erRo0ePsrBAhw4d+PTTT5kzZw6zZ8/mlltu4d5772XhwoVll+SvvPJKWBVYgPr168dlh79/SKhD45eWljJ48GAeeuih/fZr3bp1XMeJl8pmC/wDmArcDNznFQccqqq3JtWqNKEI6AcU4xJnz8FqamUyBQWpk7fXX6Dwxhtv3M8z++6772jatGm1jvHhhx+yc+fOsgqvCxYsICcnh8MOO4xvvvmGZcuW8fjjj3P66acD8P7777Nv376wMRo0aMDgwYMZPHgwd955J23atOHtt98mPz+f3Nxcvvzyy7DaWonm2GOP5YUXXuCQQw6JW7SrS2WzBfJU9c/AXgBV/QhI3ieRZhTihLXEey2sTWOMpBMK+YwdmxqlfCZMmICq0rt3b/7xj3/w6aef8sknn/CnP/0p7FJ5w4YNLFmyhM8++wxwd8+XLFnCli1bKhx/3759XHXVVXz88ce8/vrr3HnnnVx99dUccMABNGvWjBYtWjB58mS++OIL3nzzTa677rqymQwATz/9NE8++SQffvghK1euZMqUKdSvX5+uXbvSuHFjbr31Vm699VaeeuopvvjiC5YsWcITTzwRFgapLtdffz1bt27loosuYuHChaxYsYLZs2dzzTXXsH379soHqAaVea4bReRgXEGzEA0AROQPwBLgf8BHqro7wv4ZTQHOYw15rgW1aYxRI6RSyKdz5868//773H///dxxxx2sXbuW5s2bc/TRRzNx4sSyfk888QT33ntv2frgwYMBmDJlCkOHDo06ft++fTnyyCM5/fTT2bVrFxdeeCHjx48HICsri+eff54bb7yRnj170qVLFx5++GEuvPDCsv2bNm3KAw88wK233srevXvp0aMHL730Ep07dwZg7NixtG7dmoceeohhw4Zx4IEHcswxx3D77bcn7DNq164db7/9NqNGjWLgwIF8//33dOzYkQEDBpCbm5uw40SiwkoEItIFeAI4ArgNGAA0UdULRGQwcBRwNNDT2+VjnNj+T1VfSabhiaQ6lQiszHZ6YJUIjHhIeiUCVf1CRH4InAf0AhYBU7xtM4CyCWne01s9cWJ7BpA24lod8okuqvaklmHUXSp9/FVVi4EXvKWifnuAxd5S56mNaTuGYaQOsTz+ukJEnhaRn4vI4TVhVCZgT2oZRt0mlsQtnYBDgMsBRGQTMN9b5qnqB9F3rbuk0rQdwzBqnljEdSpwGtDZW28NXOgtiMg24C1VTY88YDWEPallGHWbWGKuQwFEpB3QF+jjvR7hdWkC/DBJ9qU1qTRtxzCMmiXmfK6qug74m4i8hvNkLwQujWcMwzCMukKlwigibXDe6mne65GAeEsJ8F/gzSTaaBiGkXbE4nWuo/wJrb3AQmCet7ylqsl9hixDsDmvhlG3iKeG1ibg18CNwC9V9TUT1tiw6gRGumEFCqtPLOI6DSesrYGxwLvAdyLymoiMEpFTRKRm082kGTbn1agKiShQuGXLFm644QaOOOII8vLy6NChA8OGDeObb76prdOqFhlVoDCUMNsr4RKaLXAqcBYu1wDAbmD/bLQGYHNejapT3QKF69atY+3atYwfP54ePXqwdu1ahg8fziWXXMKsWbNq45SqRToVKIy7JArQEDgTl2OgGCgFSuIdJ5WWRJZ5iUZNlgcx9idRZV5qkiuuuEIHDx4cdfugQYO0bdu2YaVYQmzZsiXqfjNmzFAR0a1bt1Z67LFjx2qrVq30gAMO0KFDh+quXbvK+rz22mt66qmnatOmTbVZs2Y6YMCA/T7ne++9Vzt27Kg5OTnaunVrvfzyy8u2lZaW6gMPPKCHHnqoNmjQQHv27KnPPvts2P7EUOblxRdf1P79+2teXp52795dZ82aFTbGxx9/rD/84Q+1UaNG2rJlS7344ot1/fr1Uc9dNTFlXmJ5/LWJiAwWkQdEpAj4FpgJDAGyk6L4GUh+Powa5d6PG2dxV6N6VKdA4bZt28jNzaVhw4YVHsMKFFaTytQXN0OgxFtKfcv3uEdgfwMMiEXJU3WpCc9V1YoX1iYJK1C4+h29f979+s7q5P/nJbpAoarqt99+q126dNEbbrih0mNbgcLIkIgChR4h73QnsIDyaVgL1GXCMmKkNmowGYmjaE0R/ab2o7ikmJzsHOYMmUN+h+T+ByayQOHOnTvL6l6Fkl5XhBUorB6xiOsdODFdpKolSbOkDmA3ttKbwlWFFJcUU6IlFJcUU7iqMOnimqgChTt27OCHP3RPqU+fPp0GDRpU2zYrUFgxlVV/7ayqD8YykLizOlhV1yTEsgzEn8ylefPyKVnmvaYHBZ0KyMnOKfNcCzoV1Jot8RQo3L59O4MGDUJVmTlzZkQvLhJWoLB6VOa5FonIDOBJVY14C0ZEmgEX4x4ueBz4Y2JNzCxCQmqJtNOP/A75zBkyh8JVhRR0Kki611oZEyZM4OSTT6Z3796MHTuWo446ClVl7ty5jBs3jtWrV7N9+3YGDBjAtm3bePnll9m5cyc7d+4EnEDn5OREHT9UoPDuu+9m3bp1YQUK8/LyygoUdujQgbVr13LbbbftV6Bw3759nHjiiTRq1Ijnn38+YoFCVaVPnz7s2LGDBQsWkJWVxTXXXJOQz+j6669n8uTJXHTRRdxxxx20bNmSFStW8MILL/Dwww/TuHHjhBwnEpWJ6xHAXcAMESnBVRlYj7uZ1QzoAXTHPVgwUlX/kzRLMwiLvaYv+R3ya11UQ8RSoHDx4sUsWLAAgMMPD891P3fuXAoqiE1ZgcLqUWGBwrJOInnAYNzDA4cAecDXuKQt/1FXcjttqU6BwqpgJWBqHitQaMRD0gsUhlBXNvtFbzEiEE8V2GAibXBzXy2pi2FkDpaLNQEUAf1wj6vlAHOITWDz88O92OxsuOoqGDLERNYw0p14smLth4hME5EDvffni8gl4p+0VkcoxAlrifdaGM++heHx14kTLXOWYWQC1RJXoIuqbhORnriMWSdRB2cLFOA81mzvtSCefQtc3DX0k6RqmbMMIxOobligRESygCuA36rqcyKyOAF2pRX5uFBAIbHFXMP29eKvU6fClCmwb58T2+bNXRy2eXP45huLxyYCVaUOXlgZcVKVp98iUV1xnQC8jytSeK/XdkA1x0xL8olPVMP29eKvQ4aUP2AwciTs2QOlpZCVBbm58MgjJrRVJTs7m71791Y4r9MwAHbv3p2QBw6qFRZQ1SdxzlovVd0hIl1w+QfiQkT6iMi/RWStiKiIDA1sf9pr9y8LAn1yReQPIvK1iOz0xju4GqdX44QyZ33zjQsNhJL2lJY6oR0xwlUzKCiAYcNg0iTLsBUrTZs2ZePGjcnPhGSkLarKrl27WLt2La1atar2eImYLdBCVb8A8F6HVmGMRsBHwFRvicRs4HLfenFg+yPAucAlwDfA74DpInJcuuVECMVh/Z5rVpa76VVa6l6feML1DXq1/jACWN2uEC1atOCrr77i008/rW1TjBSmfv36tG7dmgMPPLDaYyVCXI8SkfaqWuUKsKr6KvAqOC81Src9qroh0gYRaQL8HLhSVV/32i4HvgT6A2n15FgwB0FINEeOhO+/dze9Qvi92pD4ZmVBvXruJtm+feFTvCB8XL8I+4+VaQKdlZW1X4IQw0gmMYmriBQAf8c99nqbqv7Dt/mfuHs5fRNtXIBTRWQT8B2ulPddqrrJ23YcUB8oq1uhqmtEZBlwMmkmrlAeh/XTq1f5ja+9eyN7teBe9+5171XdtokT4amnnOD69w2JcKhNxO2TDIGuSls6C7pRt4nVc70NeAv4DJggIp+p6gfetrOBY5JhnI+ZwEvASqATcB/whnfJvwdog5tm+nVgv43etowg0o0vv1frDyOEhLG42Imlarjgwv4iXNG26gp0Rf0raotH0GNtS5UxzLbUsC1ZP+KxiuuBqjoYQEReAW4TkceAvwCHUj5TICmo6t99qx96072+xOU7eKmCXQWIOK9CRK4BrgHS7nIxmlcb6cvjn+KVnV01YUyUQMfbFnqoIhZBj7UtEcKfyB8Ps612bQtNfUxGfo9YxfW70BtVLRKRn+PCBMuBq1R1fmLNqhhVXSciXwFdvaYNuDn8LYDNvq6tcIm+I40xCZgELnFL8qytGSIJbqg95OlW1QtIhEBX9Q8mVkGPtS1VxjDbUmfcZGWmq1RcRaQ+0ENETgPe8mrIzMfdYLo+sebEhoi0ANrj0h+CS4W4F1eV9q9en4Nx6RDfqQ0bU4mg8EYT4Yr2r65Ax3up99//xifodd0DM9uq57mGvn+JpNKUgyKSjRMuBbYA03Hi+pqqrq9o35iNEGkEhGpZvAP8Fvi3d7wtwBhgGk5MOwHjgA5Ad1Xd7o3xJ+Ac3NNioalYzYBKp2LVdMpBIzaKiuIT9LoeOzTbaibmGmvKwVjzua4GbgFOw93AOgQ3c+B14E+qOjN20yKOXwDMjbDpGWAY8DLwA6ApTmDnAqP9JWVEpAHwIHApLt/sHGB4LGVnTFwNw4iVRIvrX1X1Ut/6UbgJ+2cDxwIjVPWJathbq5i4GoYRK7GKa1aM400UkbNDK6r6P1Udq6on4C7nu0bf1TAMo+4Rk7h6T1+tiLJtFXB3Am0yDMNIe2L1XFHVjyvYtjMx5hiGYWQGMYurETtFuOkMlqzKMOouVkMrwVSlnpZhGJmHea4JppCq19MyDCNzMHFNMAVUvZ6WYRiZQ8ziKiJXV7Atbee4JppQPa2xWEjAMOoy8cRcx4vIFlWd5m8UkUnAWYk1K72pTj0twzAyg3jCAj8GnhKRfqEGT1gHAqcn2jDDMIx0Jp55rnOAq4AXReREEZmM81gLVDXiAwaGYRh1lbimYqnqNBE5CJcjdT3Q13tCyzAMw/BRobh61QYisRH4EPg/EQFAVW9MrGmGYRjpS2Wea68o7ctx5bCjbTcMw6jTVCiuqmo3qgzDMKpAXDFXEbkI93RnK8JvhqmqnptIwwzDMNKZmMVVRB4ERuKqAKwjSlVVwzAMIz7PdQhwiaq+mCxjDMMwMoV4HiLIApYky5BMxFIPGkbdJR5xnQT8LFmGZBqh1IOjvVcTWMOoW8QTFmgKXCoiZwL/w5XbLsPmuYZTyP6pBy3fgGHUHeIR1x6UhwWOCGyzm1sBCnApB0NJswtq0xjDMGqcmMXV5rzGRyj1YCFOWM1rNYy6hZV5SSKWetAw6i7xPkRQDzgB6Ii72i1DVacm0C7DMIy0Jp6HCI4AXgE6A4K7V1MPd2NrD2DiahiG4RHPVKxHgMVAE2AX0B3ojbvJdWHiTTMMw0hf4gkLHI/L37pTREqBeqr6vojcDvwBOCopFhqGYaQh8XiugvNYATYD7b33XwFdEmmUYRhGuhOP5/oRcDSwAngXuENESoCrgS+SYJthGEbaEo+4/gY4wHv/K2A6LkPW18BPE2xXxlGEzXk1jLpEPA8R/Mf3fgXQw6un9a2q2hNaFRDKMxB6WmsOJrCGkenEE3PdD1XdoqoqIh0SZVAmUsj+eQYMw8hsqiWuItJGRB4HPkuQPRlJAc5jzcbyDBhGXaFScRWRpiLyFxHZLCLrRORGcdyDu7l1AnBV0i1NY0J5BsZiIQHDqCvEEnO9H+gDPAMMBH4PnIm7uTVIVd9MnnmZg+UZMIy6RSziOhi4UlVni8gE3LSr5ao6MrmmGYZhpC+xxFzbAUuhbJbA98DkZBqVyVjpF8OoG8TiuWYRXnWghPIntYw4sClZhlF3iEVcBXhORPZ46w2AySISJrCqek6ijcs0CrHSL4ZRV4hFXJ8JrD+XDEPqAgVY6RfDqCtUKq6qemVNGFIXsNIvhlF3sDIvNYxNyTKMukG1ntAyqo7NGjCMzMY811rAZg0YRuZjnmstUIglcjGMTMfEtRYowBK5GEamE5e4isggEZkuIktDaQZF5Bci0i855mUmlsjFMDKfmMVVRC4DXgA+x5XXru9tygZuT7xpmU0+MAoTVsPIVOLxXG8HrlbVm4F9vvYFwDEJtaqOYTMHDCPziEdcuxL5738HcGB1jBCRPiLybxFZKyIqIkMD20VExnj5ZHeLSKGIHBno00xEnhWRrd7yrIg0rY5dNUFo5sBo79UE1jAyg3jEdR1weIT2PsDyatrRCFdd9iZgd4TttwO3ADcAxwObgNdFpLGvz1+BY4FBuLyzxwLPVtOupFOIzRwwjEwknnmuk4DHROQX3noHETkNGA+MqY4Rqvoq8CqAiDzt3yYiAowEfquq07y2K3ACeykwUUS64wT1VFV9x+tzLTBfRLqp6qfVsS+ZFGD5BgwjE4mn+ut4EWkCvI7LjDUX2AM8pKqPJ8k+cDfP2gCzfLbsFpF5wMnARNx9oR3AO7793gZ2en1SVlz9+QaaU+652o0uw0hv4npCS1XvEpHfAD1wIYWlqrojKZaV08Z73Rho3wi09/XZ7C/x7VWl3eTbPwwRuQa4BqBjx44JNTheQkJqT20ZRuYQ90MEqrpLVRep6rs1IKxhhw6sS6AtuD1Sn/LOqpNUtbeq9m7ZsmWCTKw6hVjs1TAyiQo9VxH5d6wDJTFZ9gbvtQ2wxtfeinJvdgPQSkQk5L16sdqW7O/xpiQFWOzVMDKJysIC39SIFRWzEieeZwLvAYhIA+A04DavTxFuxkE+5XHXfFyFWn8cNmWxXK+GkVlUKK41lShbRBoBXbzVLKCjiBwDbFHV1SLyCHCXiHwCfAb8CncD66+enctEZCZu5sDVuHDARGB6Ks8UCOLP9VqECa1hpDNVSjnoiSEJjLn2xs0+CHGvtzwDDMVN98oDHgeaAQuBAaq63bfPZcBjlM8q+DcwIkH21SiWktAw0p94E7eMFJHVwFZgq4isEZGbvfhmlVHVQlWVCMtQb7uq6hhVbauqDVS1r6p+FBhji6r+TFUP9Jafqep31bGrtiik/ObWHtwkYntyyzDSi5g9VxEZj5u69CDlf+v5wN1AWyx5S8IowHmse4BSYDYwH/NgDSOdiMdz/QXwC1X9jaq+4S2/Aa4Gfp4c8+omoZtb/XH/QaXY9CzDSDfinef6vyhtlnQ7weTjwgG5lCfVbo5lzzKMdCGeG1pTgetxyVX8DCMNEqSkI8FHY0diN7kMI12o7CGCxwJ9fyYiZ+FyuAKcCLQD/pIc84zQ9Kxx7H+TawwmsIaRqlTmufYKrC/2Xg/xXjd4yxGJNMrYnwLsJpdhpBOVPURwek0ZYlRMKEQwBies/ptcJq6GkXrE9RCBiNQDTgA64hypEKqqFndNMqGbXPNxwpoNrMbd4DKBNYzUIp55rkcAr+Dyqwou/FcP2Iu7WjVxrQFCHuxUYAowGfcYm4UHDCO1iGcK1SO4mGsTYBfQHffY6hLgwsSbZkQjH3fpsI/yFIVTsWlahpFKxBMWOB7oq6o7RaQUqKeq74vI7cAfgKOSYqERkQLKUxRm47zYfdg0LcNIFeLxXAXnsQJsprwKwFeUZ7QyaohQeGAscBXlXqzlIjCM1CAecf0IONp7/y5wh4j0xWWv+iLRhhmVkw+MAobgPNbQo7KzsTLdhlHbxCOuv8F5r+DyqXbApQkcANyYYLuMOIiWi8DisIZRe8RT/fU/vvcrgB4ichDwrb8woFE7RJqmFYrDZuNCB0OwWKxh1BQxe64i0jGYt9XLoaoiUrvlUw0gehy2GFeWwUIFhlFzxBMWWIkr+BeGiDT3thkpQDAOG/o1VCxUYBg1STxTsaKVqW4EfJ8Yc4xEEXzYIBQesFCBYdQMlYqrLzOWAuNEZJdvczbucdglSbDNqCahjFpDcDkIVuOe6Crxlom4p7sewZX5LcCE1jASRSyeaygzluCeyir2bSsG3gceSrBdRgIJiWwRTky/x/1SKm5e7AjcDAPzZg0jcVQqrqHMWCIyBbhJVbd56/WABgmsAGskmUihglCSiFLKvdmnMJE1jOpS6Q0tEeknIj9V1St9wnonsAP4TkRmikjTZBtqJIZ84E+4CcpjcbXKc9n/xldodsEk7AaYYVSFWMICo4BXQysicgJwP/BnYBlwG3CX92qkCaFQAbi4T8ibLcZCBoaRCGIR157AHb71nwDvqOrVACKyBrgPE9e0xX/jq7KQgf8GWHPsRphhRCMWcW0KbPKtn4LPkwXeozyJi5HGBGcXhIoiRroBFhLdLFxYwWYcGEY4sYjreuAwYI2I5AI/AEb7tjfG/c0ZGUK0kEHQm8V7jRQ++AEmtkbdJhZxfQ0Y793EOgfYiXuEPcRRWFasjCWaNxsqlJjlLf7wwRPevkGv1sIIRl0iFnG9G3gJl8luB3CFqvrnul4FvJ4E24wUIujNFlIulsHwQQi/VxstjGCCa2Qqscxz/RroIyJNgB2qWhLo8hOc6Bp1BL/QhvCHD/YS2auF+ATXhNdIZ+JJObg1SvuWxJljpCuRwgd+rzZaGAEiC24okUVFni7esQow8TVSj7hKaxtGZUTzaguJT3BD4YVonm49nAAHk9AQOFaBr60ggm2GkSxMXKvApDWTmLZqGhd2upBrOlxT2+akPFURXL/nGsnT3eu9V8If2xXCwxKRRNg/kwGii7EJtFEdTFzjZNKaSVw79VoogVnZs1g+cDlNdzWlecPmfLPrGwo6FZDfwf78KiNWwY3m6YZE0/9EmV9wIbII+2cyhMaIJMbJFGiLJdcNTFzjZNqqae6vVIF98NCrD0EplFJKlmSRm53LIwMfMaGtApEEN0RQeAu89mC+2mjCGBLhEEHhjbWtugLtb6vsJh5RzrmytkSMUYAJf3URK38FvXv31kWLFsXU1++5IiAq+D/DLLLIzsqmVEvJyc5hzpA5JrBJpojyS3aoWISrIoLRBBrCE97E25aF+0GIdBOvqqKdiDHSQfgTOW4B8f2QiMhiVe1daT8T1/jEFcpjri0btuQvM//i3CYgS7LIkixKtZRSLSWLLPof2p8xBWNMYFOAkAhX9Q8xEQIdnKZWSnks2U91RDsRY6Sy8Cdy3H24ckhziF1gTVzjIF5xDTGOcfxqza8oXVWKNBTO3HUmFza8kJEzR7Jn356wUIF5sJlBdQXa31bRTbzaFrBUFv5Ej5uNS785KsJ4kYhVXC3mWg0KKCC3Qy7FHYrJJptDOZRe9GJOqzmMKRzD7JWzKdVSikuKmfrBVApXFVocNs2JFheualu0m3gF3vbCKrRluvAnw3MNfQaJxDxXqu65AhRRxFSmMoUp7GMfOeQwhzmwBvpN7UdxSTHZWdkIwr7SfRaHNdKGoJeeKsKf6HELsJhr0qiOuIILD4xmNCWUkE02YxnLKEZRtKaIwlWFrN66msnvT6ZESywOaxhpTqzimlUTxmQ6BRSQQw7ZZJNDDgXe72J+h3xGnTaKIUcPISc7hyyyKKWU2Stn029qP4rWWPEUw8hUTFwTQD75zGEOYxnrQgI4b7bIqzyV3yGfOUPm0P/Q/mWzCfbs28OYwjEmsIaRoVhYgOqHBfwUUUQ/+lFMcVn8Nd+L6BStKaLf1H77zSSwhw4MI32wsEAtUUghxRRTQgnFFFNIYdm2aB7siFdHMHruaAsVGEYGYeKaYPzx12yyWc3qsvAAOIEdUzCG3OxcsiWbrKwsSrSEEi2xUIFhZBAWFiCxYQGIPj0r3zfhIzSToHnD5hEfOrBQgWGkJvYQQS2STz6FFLKPfWHhAb+45nfILxPNXq16hT10EAoVlGop2VnZXHXMVQw52mUrtQcRDCM9MM+VxHuusP+NrUd4hG/4hgIKwkS2rL93s6u4pBgRKctPACAI9bPrlz2IEBLcH7T9gXm3hlHD2EMEcZAMcQUnsIUU0pzmjGRkxBkEYf0DoYLv932Pek9Ei/c0tAaeqA6GEfx5ZYGy8UyEDSMxmLjGQbLENYT/Ca4ssuhPf8YwJqLAhihaU8TUD6YyZcmUMm9VEIpLivcXWC/NYUlpSVnctl5WPQRhb8neiLHcikS4MoGuqM0E3Mh0MkpcRWQMcE+geaOqtvG2i7f9GqAZsBC4XlU/jmX8ZItrKESwB++mFVnkkhvVgw3b1/NmQwIWEly/aPrTHIaI5OnGIsKCoGhUga6oLVrIAuIT6ESIfLLHTSXb7MesZslEcb2Y8OQ1Jaq62dt+B/ArYCjwKXA3cCrQTVW3VzZ+ssUVnMCOYQyzmV0msLF4sBHH8oUPQn+QwRkHkUQwVhGuaFusbUCVBDoRIp/scVPJtkhXI6kk/MkYo7avtjJRXH+sqj0jbBNgHfBHVf2N15YHbAJuVdWJlY1fE+IK1fNgKx07ILjRvoCViXB1hCNSyKK6Al3dMZL545EKtgWvRlJJ+FP5R8l/tRVvprpMnIp1qIisxVXbWAj8UlVXAJ2BNsCsUEdV3S0i84CTcYVBU4JQDgK/B1tMMVOZSiGFUWcSxDS2b2pXsN1Pr1a9kuYFRApZZNIfYiraliXuIZRS3NVISHhLtZS9JXsT1pYqYyRj3OKSYgpXFSY8vIOomPIAAAwGSURBVJIunusgoDHwCdAKFwI4AjgS6Aa8DRyiqqt9+zwFtFfVs6KMeQ0uRkvHjh2P+/LLL5N6Dn7807Sy8XK9so9ssrmKqxjCkGp7srVFLB50Ol9CppptwauRVBL+VP5RqgnPNS3ENYiINAJWAL8FFuDEtaOqrvH1mQK0VdWBlY1XU2EBP6FpWqtZzWQmU0IJ4C79GtAgIaECo24QKQafCsKfyj9KFnOtABGZi/NkHwSWAyeo6nu+7TOAr1X1isrGqg1xDRHyYr+nfE5rNtlczdV0pGO1QgWGYSSejM6KJSINcGGB9cBKYANwZmD7acA7tWJgHITisNdyLbnkliV8mcIURjOaAgoYxrCw5C+GYaQ+aeG5ishDwCvAalzMdTTQB+ilql96U7Huwk3F+gwXk+1DCk3FioXKQgWVPUJrGEbyybTZAgcDfwNaAJtxcdaTVDV0F2o8kAc8TvlDBANiEdZUIt/7V0QRz/BMWahAUfawhxGMoJTSjLjxZRiZTlp4rskmVTxXP8G0hYJQ6v0DL5kL9U1kDaOGyfgbWokkFcU1RDD5i//GF+wfMmhOcwsdGEYSMXGNg1QWVz9+b7aY8qehssgim2xKKAl78ssE1zASj4lrHKSLuIaoLGQAJriGkSxMXOMg3cQ1RDBk4M9ZkEWWCa5hJAET1zhIV3H14xfakFBWVXDrUa/sRhkQNq4Jr1HXMXGNg0wQ10hUVXChfDaCIOxlb6WeLpgIG3UDE9c4yFRxjURlghtKggEVpLiL4OkGRbgy7zeWNhNqIxUxcY2DuiSukfAL7n/5b9mNslDGLr9oRvJ0o+Uljeb9RhNjf1tZpqM4veVY2szjNqqDiWsc1HVxDRIS22jCFAwtRBNGSEzi6Hi85XhEO1L/WIU8ESKfrHHNtvjGiPcH1cQ1Dkxc4yMYWgh+eSvzfuMVwVi95URk+49FyCsT6GQJf02NUZds28c+KqrIHAkT1zgwcU08FXm/8XojsXjLifpDjCfsEW1bsoS/JseoS7Zlk81YxjKKUfuNFwkT1zgwcU19KvOWE3UJGU/YI5U8MLPNPNeUxMTV8BOLkKdi7NBss5hrymHiahhGrGR0JQLDMIxUx8TVMAwjCZi4GoZhJAETV8MwjCRg4moYhpEETFwNwzCSgImrYRhGEjBxNQzDSAImroZhGEnAntACRGQz8GWcu7UAvk6COTWJnUNqkO7nkO72Q3zncIiqtqysk4lrFRGRRbE8ApfK2DmkBul+DuluPyTnHCwsYBiGkQRMXA3DMJKAiWvVmVTbBiQAO4fUIN3PId3thyScg8VcDcMwkoB5roZhGEnAxNUwDCMJmLhWAREZLiIrReR7EVksIqfVtk2REJFRIvKeiGwTkc0i8oqI9Az0EREZIyLrRGS3iBSKyJG1ZXNliMgvRURF5I++tpQ/BxFpKyLPeP8P34vIUhHp69uesucgItkiMtb3nV8pIveJSD1fn5SyX0T6iMi/RWSt930ZGtheqb0i0kxEnhWRrd7yrIg0jdkIVbUljgW4CNgLXA10B/4A7AA61rZtEWz9D3Al0BPoBfwT2AAc5OtzB7AduNDr9wKwDmhc2/ZHOJ+TgJXAB8Af0+UcgKbACmAqcALQGegHdE+HcwB+CWwBzgY6AecA3wKjU9V+4IfA/cCPgV3A0MD2Su0FXgM+Bk4G8r33r8RsQ23/x6XbAiwEJgfaPgfG1bZtMdjeCCgBzvbWBVgP3OXrk+d96a6tbXsDtjcBlgNnAIUhcU2Hc/D+yN+uYHtKnwMwHXgm0PYMMD1N7N/hF9dY7MU5Tgqc4utzqtfWLZbjWlggDkQkBzgOmBXYNAv365bqNMaFgr711jsDbfCdj6ruBuaReuczCXhRVd8ItKfDOZwHLBSR50Vkk4gsEZERIiLe9lQ/h7eA00XkCAAR6YH7kXvV257q9geJxd58nCi/49vvbWAnMZ5Tvcq7GD5aANnAxkD7RqB/zZsTN48CS4Aib72N9xrpfNrXlFGVISJXA12AyyNsTodzOBQYDvwe+C1wDC6cBPBHUv8cHsD9MC8VkRKcbvxGVSd421Pd/iCx2NsG2Kyeywqgqioim3z7V4iJa9UITg6WCG0phYj8DndZc6qqlgQ2p+z5iEg33GX1aapaXEHXlD0H3NXCIlUd5a3/V0S6AtfjxDVEqp7DRcAQ4FJc3PEY4FERWamqf/b1S1X7o1GZvZFsj/mcLCwQH1/jYpbBX65W7P8rmDKIyO+BS4AzVHWFb9MG7zWVzycfd8XwkYjsE5F9QF9guPf+G69fKp/DemBpoG0Z0NF7n+r/Dw8CD6nq31X1Q1V9FvgdEPqxSHX7g8Ri7waglS90g/e+JTGek4lrHHie02LgzMCmMwmPzaQMIvIozuM4Q1U/CWxeifsSnenr3wA4jdQ5n5dxMx2O8S2LgL977z8j9c/hbaBboO1wytNcpvr/Q0OcU+GnhHL9SHX7g8RibxHuBnC+b7984ABiPafavpOXbgvuEqkY+AXujuKjuMD3IbVtWwRbHwe24W4+tPEtjXx97vD6XPD/7d1diBV1GMfx76+ITOhNCHqBIjSjNDWyKMFVKAgSgl6MIikqLCqLom4iELso6kLwLiKyTTQMQ1ASIjZMLzIqkjQzbJECpURb32htSXq6eGZhGPflrNt05sjvA3+YmfP/73lmd32cnfmf509OSVlLQ6YAjXBeX3DqVKzGngNwMzl971Xy3vFC4CjwbCecA9AN7AMWkFOx7gEOAsubGn+RGAf/M+4HlhbbV7YaLzkVayc5BfC2YttTsWr+wT0D/AIMkFeyXe2OaZg4Y5i2rNRHwDLyT9e/gC3A9HbHPsp5VZNr48+hSEzfF/HtAZ6nqO3R9HMgH2atIK+0T5Bzdt8AJjQ1fmD+ML/73a3GC0wCVhdJ+FixfVGrMbhwi5lZDXzP1cysBk6uZmY1cHI1M6uBk6uZWQ2cXM3MauDkamZWAydXs1EUxZbvP41x50jaI6lrDGOWSNo41vey5nFytcaS1F0ktmr7qt2xtehJYH9EbB3DmHeB2U1d3cJa5+RqTdcDXFZpd7U1otY9B7w3aq+SiBgAPiQ/wWUdzMnVmm4gIn6vtL7BF4sr2SWSNknql/SrpEXlLyDpBkk9xVpJfcUV8YWVPo9K2ilpQNIBSd2VOCZJWifpT0l7q+9RJWk2WZzlk8rxKyStlXS4aJuK8oNlG4G7JU1s7VtkTeTkameC18iENItcsWBVkdwoEtSnZHGdW8iiI3OAlYODJT0FvAO8D8wgr4x3Vd5jKbABmAl8BKyUdNUIMc0FeiPiSOl9JgKbyc+yzyOLgfwG9FQS6bdkreVyRSbrNO0uCuHmNlwjqzGdJBNjub1V6hOcuqZZD7C62F5MVqAqVzuaX4ybUuzvA94cIY6gtEYamfj6gUUjjFkBbKkce5xcb61csOVssibtA5W+fcAT7f4ZuJ1+80oE1nRbyQdDZUcq+9uG2F9QbF8H7IiI46XXvwT+Aa6XdIxc2uPzUeLYMbgRESclHSSLKw/nPPIKtewmcv2m46UazJD1UidX+p4ovoZ1KCdXa7r+iOgdx/iRluWI4vVW/D3E2JFuqx0CbqwcO4tcw+zBIfr3VfYnkTVTrUP5nqudCW4dYn93sf0jMFPS+aXX55C/+7sj4gCwH7j9P45pO3CtpPK/se/IYtmHIqK30soP6SYDE4r+1qGcXK3pzpV0aaVdUulzr6TFkq6R9AqZKFcUr60hl0NeVcwa6CIfXq0vXRG/Drwg6UVJUyXNkvTSOOPeTCbIGaVja8j1lzZImifpakldkpZXZgzMBfZGxM/jjMHayMnVmu4O8ol6uW2v9FkG3EfeF30aeCwivgGIiH7gTuAC4Gvyif828uESRZ+3yZVYFwM/kLMLpo0n6Ij4A1gPPFw61g90kZX81wE/AR8AFwOHS8MfIj9MYB3MKxFYR5MUwMKI+LjdsVRJmkZewU6JiGMtjplOPlybGhFH64zP6uUrV7OaRMQu4GVyhkCrLgcecWLtfJ4tYFajiFg1xv6f1RWL/b98W8DMrAa+LWBmVgMnVzOzGji5mpnVwMnVzKwGTq5mZjVwcjUzq8G/zr4T/pks8IsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(range(0,len(fc1_ranks)), fc1_ranks, marker='.', color='cyan', label=\"FC1 dropout\")\n",
    "plt.scatter(range(0,len(fc2_ranks)), fc2_ranks, marker='.', color='lime', label=\"FC2 dropout\")\n",
    "plt.scatter(range(0,len(fc1_ranks)), fc1_bsln_ranks, marker='.', color='blue', label=\"FC1 baseline\")\n",
    "plt.scatter(range(0,len(fc2_ranks)), fc2_bsln_ranks, marker='.', color='green', label=\"FC2 baseline\")\n",
    "\n",
    "#lt.ylim([0.905,0.920])\n",
    "plt.title('Mini-AlexNet w/Dropout:'+'\\n'+'Stable Rank vs Epoch')\n",
    "plt.xlabel(\"Epoch (e)\")\n",
    "plt.ylabel(r\"Stable Rank $\\mathcal{R}_{s}(\\mathbf{W}^{e}_{l})$\")\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(RANK_FILE)\n",
    "print(\"saved file {}\".format(RANK_FILE))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Read Dropout Training accuracies from output file\n",
    "\n",
    "Notice: there is no epoch=0 training accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:58:42.807592Z",
     "start_time": "2018-09-02T22:58:42.416093Z"
    }
   },
   "outputs": [],
   "source": [
    "dp_train_accs = RMT_Util.read_training_accuracies(DROPOUT_outfile)\n",
    "assert(len(dp_train_accs)==NUM_DROPOUT_EPOCHS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-04-11T00:59:51.698081Z",
     "start_time": "2018-04-11T00:59:38.838Z"
    }
   },
   "source": [
    "#### Recompute Baseline Test accuracies "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:58:43.295313Z",
     "start_time": "2018-09-02T22:58:42.813826Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/charleshmartin/anaconda3/lib/python3.6/site-packages/tensorflow/python/client/session.py:1645: UserWarning: An interactive session is already active. This can cause out-of-memory errors in some cases. You must explicitly call `InteractiveSession.close()` to release resources held by the other session(s).\n",
      "  warnings.warn('An interactive session is already active. This can '\n"
     ]
    }
   ],
   "source": [
    "opt = SGD(lr=0.01, momentum=0.9)\n",
    "model.compile(loss='mse', optimizer=opt, metrics=['accuracy'])\n",
    "\n",
    "sess = tf.InteractiveSession()\n",
    "sess.run(tf.global_variables_initializer())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-02T22:58:48.654671Z",
     "start_time": "2018-09-02T22:58:43.298057Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading test images...\n"
     ]
    }
   ],
   "source": [
    "test_images, test_labels = RMT_Util.read_test_images(datadir=DATA_DIR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:10:38.887031Z",
     "start_time": "2018-09-02T22:58:48.657821Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1239392ebeb0451d807d3d68d1c9f049",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 9s 914us/step\n",
      "10000/10000 [==============================] - 9s 913us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 924us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 919us/step\n",
      "10000/10000 [==============================] - 9s 922us/step\n",
      "10000/10000 [==============================] - 9s 917us/step\n",
      "10000/10000 [==============================] - 9s 917us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 915us/step\n",
      "10000/10000 [==============================] - 9s 917us/step\n",
      "10000/10000 [==============================] - 9s 924us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 929us/step\n",
      "10000/10000 [==============================] - 9s 940us/step\n",
      "10000/10000 [==============================] - 9s 928us/step\n",
      "10000/10000 [==============================] - 9s 922us/step\n",
      "10000/10000 [==============================] - 9s 923us/step\n",
      "10000/10000 [==============================] - 9s 924us/step\n",
      "10000/10000 [==============================] - 9s 931us/step\n",
      "10000/10000 [==============================] - 9s 942us/step\n",
      "10000/10000 [==============================] - 9s 930us/step\n",
      "10000/10000 [==============================] - 9s 927us/step\n",
      "10000/10000 [==============================] - 9s 929us/step\n",
      "10000/10000 [==============================] - 9s 923us/step\n",
      "10000/10000 [==============================] - 9s 927us/step\n",
      "10000/10000 [==============================] - 9s 943us/step\n",
      "10000/10000 [==============================] - 9s 930us/step\n",
      "10000/10000 [==============================] - 9s 933us/step\n",
      "10000/10000 [==============================] - 9s 924us/step\n",
      "10000/10000 [==============================] - 9s 927us/step\n",
      "10000/10000 [==============================] - 9s 927us/step\n",
      "10000/10000 [==============================] - 9s 924us/step\n",
      "10000/10000 [==============================] - 9s 929us/step\n",
      "10000/10000 [==============================] - 9s 927us/step\n",
      "10000/10000 [==============================] - 9s 930us/step\n",
      "10000/10000 [==============================] - 9s 924us/step\n",
      "10000/10000 [==============================] - 9s 928us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 937us/step\n",
      "10000/10000 [==============================] - 9s 929us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 927us/step\n",
      "10000/10000 [==============================] - 9s 929us/step\n",
      "10000/10000 [==============================] - 9s 926us/step\n",
      "10000/10000 [==============================] - 9s 922us/step\n",
      "10000/10000 [==============================] - 9s 920us/step\n",
      "10000/10000 [==============================] - 9s 918us/step\n",
      "10000/10000 [==============================] - 9s 933us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 923us/step\n",
      "10000/10000 [==============================] - 9s 928us/step\n",
      "10000/10000 [==============================] - 9s 943us/step\n",
      "10000/10000 [==============================] - 9s 926us/step\n",
      "10000/10000 [==============================] - 9s 923us/step\n",
      "10000/10000 [==============================] - 9s 922us/step\n",
      "10000/10000 [==============================] - 9s 927us/step\n",
      "10000/10000 [==============================] - 9s 927us/step\n",
      "10000/10000 [==============================] - 9s 935us/step\n",
      "10000/10000 [==============================] - 9s 929us/step\n",
      "10000/10000 [==============================] - 9s 930us/step\n",
      "10000/10000 [==============================] - 9s 922us/step\n",
      "10000/10000 [==============================] - 9s 930us/step\n",
      "10000/10000 [==============================] - 9s 915us/step\n",
      "10000/10000 [==============================] - 9s 914us/step\n",
      "10000/10000 [==============================] - 9s 929us/step\n",
      "10000/10000 [==============================] - 9s 916us/step\n",
      "10000/10000 [==============================] - 9s 924us/step\n",
      "10000/10000 [==============================] - 9s 915us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 916us/step\n",
      "10000/10000 [==============================] - 9s 926us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 915us/step\n",
      "10000/10000 [==============================] - 9s 918us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 919us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 917us/step\n",
      "10000/10000 [==============================] - 9s 914us/step\n",
      "10000/10000 [==============================] - 9s 918us/step\n",
      "10000/10000 [==============================] - 9s 917us/step\n",
      "10000/10000 [==============================] - 9s 916us/step\n",
      "10000/10000 [==============================] - 9s 914us/step\n",
      "10000/10000 [==============================] - 9s 919us/step\n",
      "10000/10000 [==============================] - 9s 916us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 916us/step\n",
      "10000/10000 [==============================] - 9s 918us/step\n",
      "10000/10000 [==============================] - 9s 921us/step\n",
      "10000/10000 [==============================] - 9s 925us/step\n",
      "10000/10000 [==============================] - 9s 917us/step\n",
      "10000/10000 [==============================] - 9s 913us/step\n",
      "10000/10000 [==============================] - 9s 920us/step\n",
      "10000/10000 [==============================] - 9s 920us/step\n",
      "10000/10000 [==============================] - 9s 914us/step\n",
      "10000/10000 [==============================] - 9s 919us/step\n",
      "10000/10000 [==============================] - 9s 910us/step\n",
      "\n"
     ]
    }
   ],
   "source": [
    "dp_test_accs = []\n",
    "for epoch in tqdm(range(NUM_DROPOUT_EPOCHS)):\n",
    "    modelfile = DROPOUT_per_epoch_weightfile.format(epoch)\n",
    "    model.load_weights(modelfile)\n",
    "    acc =  model.evaluate(test_images, test_labels, batch_size=256)\n",
    "    dp_test_accs.append(acc[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training and test Accuracies vs Epoch\n",
    "\n",
    "## NO Overtraining w/Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:10:38.904531Z",
     "start_time": "2018-09-03T00:10:38.892634Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(100, 100)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(dp_train_accs), len(dp_test_accs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:10:39.586361Z",
     "start_time": "2018-09-03T00:10:38.907306Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "saved file img/mini-alexnet-dropout-accuracy-per-epoch.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = range(NUM_DROPOUT_EPOCHS)\n",
    "\n",
    "plt.scatter(x, bsln_train_accs,  marker='.',color='blue', label=\"baseline train\")\n",
    "plt.scatter(x, dp_train_accs,  marker='.',color='blueviolet', label='dropoout train')\n",
    "\n",
    "plt.scatter(x, bsln_test_accs,  marker='.',color='green', label=\"baseline test\")\n",
    "plt.scatter(x, dp_test_accs,  marker='.',color='teal', label='dropout test')\n",
    "\n",
    "\n",
    "plt.title('Mini-AlexNet w/Dropout:'+'\\n'+'Accuracies vs Epoch')\n",
    "plt.xlabel(\"Epoch (e)\")\n",
    "plt.ylabel(\"Accuracy\")\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(ACCURACY_FILE)\n",
    "print(\"saved file {}\".format(ACCURACY_FILE))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fit Marchenko Pastur to ESD\n",
    "\n",
    "Layer 7 and 8(9): both have a very close outlier (finite size or bleeding out ?) just near the MP edge.  This makes it hard to fit $\\sigma_{emp}^{2}$, so we compare to the ESD of shuffled X and make sure the estimate is smaller\n",
    "$$\\sigma_{emp}^{2}<\\sigma_{shuf}^{2}$$ "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  ESD of Shuffled X"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:25.438473Z",
     "start_time": "2018-09-03T00:10:39.588693Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "get_shuffled_eigenvalues\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2bc54ac133e346708525dd4baf7144a8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "Q = Q7\n",
    "model.load_weights(DROPOUT_final_weightfile)\n",
    "W7 = model.layers[LAYER_FC1].get_weights()[0]\n",
    "W7.shape\n",
    "evals = RMT_Util.get_shuffled_eigenvalues(W7, layer=LAYER_FC1)\n",
    "evals = (np.array(evals).flatten())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:26.000061Z",
     "start_time": "2018-09-03T00:12:25.441251Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "% spikes outside bulk 0.00\n",
      "% sigma 1.2538\n",
      "1.2538188753737491\n",
      "1.2538188753737491\n",
      "saved figure img/mini-alexnet-dropout-fc1-rand.png\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "RMT_Util.ipynb:7: RuntimeWarning: divide by zero encountered in true_divide\n",
      "  \"## RMT Util\\n\",\n",
      "RMT_Util.ipynb:7: RuntimeWarning: invalid value encountered in sqrt\n",
      "  \"## RMT Util\\n\",\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sigma = RMT_Util.plot_ESD_and_fit(model=None, eigenvalues=evals, Q=Q)\n",
    "print(sigma)\n",
    "\n",
    "print(sigma)\n",
    "plt.title(TITLE_RAND_FC1)\n",
    "plt.xlabel(XLABEL_FC1)\n",
    "plt.ylabel(YLABEL)\n",
    "plt.tight_layout()\n",
    "\n",
    "filename=RAND_FC1_FILENAME\n",
    "plt.savefig(filename)\n",
    "print(\"saved figure {}\".format(filename))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Various MP Fits\n",
    "\n",
    "Use in paper ?  Only if really interesting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:27.572416Z",
     "start_time": "2018-09-03T00:12:26.002440Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "RMT_Util.ipynb:7: RuntimeWarning: divide by zero encountered in true_divide\n",
      "  \"## RMT Util\\n\",\n",
      "RMT_Util.ipynb:7: RuntimeWarning: invalid value encountered in sqrt\n",
      "  \"## RMT Util\\n\",\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "% spikes outside bulk 2.40\n",
      "% sigma 1.2464\n",
      "1.2464242859936685\n",
      "saved figure img/mini-alexnet-dropout-fc1-9spikes.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Q = Q7\n",
    "sigma = RMT_Util.plot_ESD_and_fit(model=model, weightfile=DROPOUT_final_weightfile, layer=LAYER_FC1, \n",
    "                            Q=Q, num_spikes=9, sigma=None, color='blue')\n",
    "\n",
    "print(sigma)\n",
    "plt.title(TITLE_FC1)\n",
    "plt.xlabel(XLABEL_FC1)\n",
    "plt.ylabel(YLABEL)\n",
    "plt.tight_layout()\n",
    "\n",
    "filename=FC1_FILENAME_1\n",
    "plt.savefig(filename)\n",
    "print(\"saved figure {}\".format(filename))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:29.119160Z",
     "start_time": "2018-09-03T00:12:27.575219Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "RMT_Util.ipynb:7: RuntimeWarning: divide by zero encountered in true_divide\n",
      "  \"## RMT Util\\n\",\n",
      "RMT_Util.ipynb:7: RuntimeWarning: invalid value encountered in sqrt\n",
      "  \"## RMT Util\\n\",\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "% spikes outside bulk 2.67\n",
      "% sigma 1.2173\n",
      "1.2464242859936685\n",
      "saved figure img/mini-alexnet-dropout-fc1-10spikes.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Q = Q7\n",
    "RMT_Util.plot_ESD_and_fit(model=model, weightfile=DROPOUT_final_weightfile, layer=LAYER_FC1, \n",
    "                            Q=Q, num_spikes=10, sigma=None, color='blue')\n",
    "\n",
    "print(sigma)\n",
    "plt.title(TITLE_FC1)\n",
    "plt.xlabel(XLABEL_FC1)\n",
    "plt.ylabel(YLABEL)\n",
    "plt.tight_layout()\n",
    "\n",
    "filename=FC1_FILENAME_2\n",
    "plt.savefig(filename)\n",
    "print(\"saved figure {}\".format(filename))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:31.030538Z",
     "start_time": "2018-09-03T00:12:29.121521Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "get_shuffled_eigenvalues\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5c3939ff9f274fa39422e9ab282db652",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "Q = Q8\n",
    "model.load_weights(DROPOUT_final_weightfile)\n",
    "W8 = model.layers[LAYER_FC2].get_weights()[0]\n",
    "W8.shape\n",
    "evals = RMT_Util.get_shuffled_eigenvalues(W8, layer=LAYER_FC2)\n",
    "evals = (np.array(evals).flatten())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:31.485257Z",
     "start_time": "2018-09-03T00:12:31.033598Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "% spikes outside bulk 0.00\n",
      "% sigma 1.0345\n",
      "1.0344700315841424\n",
      "1.0344700315841424\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "RMT_Util.ipynb:7: RuntimeWarning: divide by zero encountered in true_divide\n",
      "  \"## RMT Util\\n\",\n",
      "RMT_Util.ipynb:7: RuntimeWarning: invalid value encountered in sqrt\n",
      "  \"## RMT Util\\n\",\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "saved figure img/mini-alexnet-dropout-fc2-rand.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sigma = RMT_Util.plot_ESD_and_fit(model=None, eigenvalues=evals, Q=Q)\n",
    "print(sigma)\n",
    "\n",
    "print(sigma)\n",
    "plt.title(TITLE_RAND_FC2)\n",
    "plt.xlabel(XLABEL_FC2)\n",
    "plt.ylabel(YLABEL)\n",
    "plt.tight_layout()\n",
    "\n",
    "filename=RAND_FC2_FILENAME\n",
    "plt.savefig(filename)\n",
    "print(\"saved figure {}\".format(filename))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:32.181833Z",
     "start_time": "2018-09-03T00:12:31.488085Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "RMT_Util.ipynb:7: RuntimeWarning: divide by zero encountered in true_divide\n",
      "  \"## RMT Util\\n\",\n",
      "RMT_Util.ipynb:7: RuntimeWarning: invalid value encountered in sqrt\n",
      "  \"## RMT Util\\n\",\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "% spikes outside bulk 4.92\n",
      "% sigma 0.9081\n",
      "1.0344700315841424\n",
      "saved figure img/mini-alexnet-dropout-fc2-9spikes.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Q = Q8\n",
    "RMT_Util.plot_ESD_and_fit(model = model, weightfile = DROPOUT_final_weightfile, \n",
    "                          layer=LAYER_FC2, Q=Q, num_spikes=9, sigma=None, color='blue')\n",
    "\n",
    "print(sigma)\n",
    "plt.title(TITLE_FC2)\n",
    "plt.xlabel(XLABEL_FC2)\n",
    "plt.ylabel(YLABEL)\n",
    "plt.tight_layout()\n",
    "\n",
    "filename=FC2_FILENAME_1\n",
    "plt.savefig(filename)\n",
    "print(\"saved figure {}\".format(filename))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:32.642570Z",
     "start_time": "2018-09-03T00:12:32.183959Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "RMT_Util.ipynb:7: RuntimeWarning: divide by zero encountered in true_divide\n",
      "  \"## RMT Util\\n\",\n",
      "RMT_Util.ipynb:7: RuntimeWarning: invalid value encountered in sqrt\n",
      "  \"## RMT Util\\n\",\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "% spikes outside bulk 5.49\n",
      "% sigma 0.8925\n",
      "1.0344700315841424\n",
      "saved figure img/mini-alexnet-dropout-fc2-10spikes.png\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Q = Q8\n",
    "RMT_Util.plot_ESD_and_fit(model = model, weightfile = DROPOUT_final_weightfile, \n",
    "                          layer=LAYER_FC2, Q=Q, num_spikes=10, sigma=None, color='blue')\n",
    "\n",
    "print(sigma)\n",
    "plt.title(TITLE_FC2)\n",
    "plt.xlabel(XLABEL_FC2)\n",
    "plt.ylabel(YLABEL)\n",
    "plt.tight_layout()\n",
    "\n",
    "filename=FC2_FILENAME_2\n",
    "plt.savefig(filename)\n",
    "print(\"saved figure {}\".format(filename))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Scree Plots:  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:33.065106Z",
     "start_time": "2018-09-03T00:12:32.644478Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "RMT_Util.scree_plot(model, DROPOUT_final_weightfile, layer=LAYER_FC1, label=\"FC1\")\n",
    "plt.title(\"Mini AlexNet w/Dropout:  Scree Plot FC1\")\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-03T00:12:33.213574Z",
     "start_time": "2018-09-03T00:12:33.067325Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "RMT_Util.scree_plot(model, DROPOUT_final_weightfile, layer=LAYER_FC2, label=\"FC2\")\n",
    "plt.title(\"Mini AlexNet w/Dropout:  Scree Plot FC2\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.5"
  },
  "widgets": {
   "state": {
    "271fdd46fa0c4eaa89f7a84a15a0a2db": {
     "views": [
      {
       "cell_index": 35
      }
     ]
    },
    "369c1bda32bc48c0aa2727942b24eba2": {
     "views": [
      {
       "cell_index": 15
      }
     ]
    },
    "4542f54dbed344279350cea482f38de2": {
     "views": [
      {
       "cell_index": 30
      }
     ]
    },
    "61e07a27c7ef422cb14d5f8ce80608bb": {
     "views": [
      {
       "cell_index": 11
      }
     ]
    },
    "9033b3a9279a43bdb48bc4fb3977d6a6": {
     "views": [
      {
       "cell_index": 24
      }
     ]
    }
   },
   "version": "1.2.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}