MLP3-Basline.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## MLP3 Baseline Results\n",
    "\n",
    "Expects:\n",
    "\n",
    "- weight.h5 files in ./weights \n",
    "- CIFAR10 data in ../data/cifar-10-batches-py/\n",
    "\n",
    "and while the keras code needs py27, the notebooks can use py3.6 \n",
    "\n",
    "### TODO:  \n",
    "\n",
    "- move weights to standard place so others can download\n",
    "- include python code & scripts to regenerate results from scratch if necessary\n",
    "- document code precisely in paper\n",
    "- test regeneration end to end with git clone\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:18.187136Z",
     "start_time": "2018-09-14T20:33:15.722931Z"
    }
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pickle, time\n",
    "from copy import deepcopy\n",
    "from shutil import copy\n",
    "import import_ipynb\n",
    "from tqdm import tqdm_notebook as tqdm\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import tensorflow as tf\n",
    "import keras\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 RMT_Util"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:18.191953Z",
     "start_time": "2018-09-14T20:33:18.188773Z"
    }
   },
   "outputs": [],
   "source": [
    "matplotlib.rcParams.update({'font.size': 16})\n",
    "plt.rcParams['figure.figsize'] = [6,5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "DataDIR = \"/Users/charleshmartin/work/DLGeneralization/data\"\n",
    "ModelDIR = \"/Users/charleshmartin/work-tmp/MLP3\"\n",
    "WeightsDIR = ModelDIR+\"/weights\"\n",
    "ResultsDIR = ModelDIR+\"/results\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:18.215917Z",
     "start_time": "2018-09-14T20:33:18.194590Z"
    }
   },
   "outputs": [],
   "source": [
    "BASELINE_outfile = ResultsDIR+'/cifar10_keras_mlp3.out'\n",
    "\n",
    "BASELINE_final_weightfile = WeightsDIR+'/model.b16.final.h5'\n",
    "BASELINE_per_epoch_weightfile = WeightsDIR+\"/model.b16.e{}.h5\"\n",
    "NUM_BASELINE_EPOCHS = 45\n",
    "\n",
    "BASELINE_init_weighfile = WeightsDIR+'/model.b16.e0.h5'\n",
    "\n",
    "REGULARIZED_final_weightfile = WeightsDIR+'/model.b16.wd.final.h5'\n",
    "RANDOMIZED_final_weightfile = WeightsDIR+'/model.b16.rand100.final.h5'\n",
    "\n",
    "LAYER_FC1 = 2\n",
    "LAYER_FC2 = 3\n",
    "\n",
    "ENTROPY_FILE = \"img/mlp3-entropies.png\"\n",
    "RANK_FILE = \"img/mlp3-ranks.png\"\n",
    "\n",
    "INIT_SEMICIRCLE_FC1_FILE =  \"img/mlp3-init-semicirle-fc1.png\"\n",
    "FINAL_SEMICIRCLE_FC1_FILE = \"img/mlp3-final-semicirle-fc1.png\"\n",
    "INIT_ESD_FC1_FILE =  \"img/mlp3-init-esd-fc1.png\"\n",
    "FINAL_ESD_FC1_FILE = \"img/mlp3-final-esd-fc1.png\"\n",
    "\n",
    "INIT_SEMICIRCLE_FC2_FILE =  \"img/mlp3-init-semicircle-fc2.png\"\n",
    "FINAL_SEMICIRCLE_FC2_FILE = \"img/mlp3-final-semicircle-fc2.png\"\n",
    "INIT_ESD_FC2_FILE =  \"img/mlp3-init-esd-fc2.png\"\n",
    "FINAL_ESD_FC2_FILE = \"img/mlp3-final-esd-fc2.png\"\n",
    "\n",
    "INIT_SCREE_FILE =  \"img/mlp3-init-scree.png\"\n",
    "FINAL_SCREE_FILE =  \"img/mlp3-final-scree.png\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:19.088272Z",
     "start_time": "2018-09-14T20:33:18.353040Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ln: model.b16.e44.h5: File exists\r\n"
     ]
    }
   ],
   "source": [
    "!cd $WeightsDIR; ln -s model.b16.final.h5 model.b16.e44.h5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cifar10_keras_mlp3.b2.out         cifar10_keras_mlp3.r100.out\r\n",
      "cifar10_keras_mlp3.b2.r100.out    cifar10_keras_mlp3.r100.wd.out\r\n",
      "cifar10_keras_mlp3.b2.r100.wd.out cifar10_keras_mlp3.wd.out\r\n",
      "cifar10_keras_mlp3.b2.wd.out      results.out\r\n",
      "cifar10_keras_mlp3.out\r\n"
     ]
    }
   ],
   "source": [
    "!ls $ResultsDIR"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Baseline "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:19.226475Z",
     "start_time": "2018-09-14T20:33:19.090899Z"
    }
   },
   "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.MLP3_model()\n",
    "\n",
    "sess = tf.InteractiveSession()\n",
    "sess.run(tf.global_variables_initializer())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:19.305002Z",
     "start_time": "2018-09-14T20:33:19.229810Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "flatten_3 (Flatten)          (None, 2352)              0         \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 512)               1204736   \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 512)               262656    \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 512)               262656    \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 1,735,178\n",
      "Trainable params: 1,735,178\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model.load_weights(BASELINE_final_weightfile)\n",
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Check baseline shapes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:19.324005Z",
     "start_time": "2018-09-14T20:33:19.307734Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2352, 512)\n",
      "(512, 512)\n",
      "(512, 512)\n",
      "(512, 10)\n"
     ]
    }
   ],
   "source": [
    "for layer in model.layers:\n",
    "    weights = layer.get_weights() \n",
    "    if len(weights)>0 :\n",
    "        W = weights[0]\n",
    "        print(W.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-04-11T00:20:35.143766Z",
     "start_time": "2018-04-11T00:20:35.141086Z"
    }
   },
   "source": [
    "## BASLINE\n",
    "\n",
    "### Compute Entropy, Rank, and Accuracies per Epoch\n",
    "\n",
    "Baseline has 44 epochs...we can do more if necessary\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:47.511629Z",
     "start_time": "2018-09-14T20:33:19.326483Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d46daf492dc641a59443e0add97d343a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=45), 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",
    "for epoch in tqdm(range(NUM_BASELINE_EPOCHS)):\n",
    "    model.load_weights(BASELINE_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": 47,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:47.520970Z",
     "start_time": "2018-09-14T20:33:47.514649Z"
    }
   },
   "outputs": [],
   "source": [
    "assert(len(fc1_entropies)==NUM_BASELINE_EPOCHS)\n",
    "assert(len(fc2_entropies)==NUM_BASELINE_EPOCHS)\n",
    "\n",
    "assert(len(fc1_ranks)==NUM_BASELINE_EPOCHS)\n",
    "assert(len(fc2_ranks)==NUM_BASELINE_EPOCHS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:47.949073Z",
     "start_time": "2018-09-14T20:33:47.523290Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(range(0,len(fc1_entropies)), fc1_entropies, marker='.', color='blue', label='FC1')\n",
    "plt.scatter(range(0,len(fc2_entropies)), fc2_entropies, marker='.', color='teal', label='FC2')\n",
    "plt.ylim([0.905,0.920])\n",
    "plt.title('MLP3:  Entropy $\\mathcal{S}(\\mathbf{W}^{e}_{l})$ vs Epoch (e)')\n",
    "plt.xlabel(\"Epoch (e)\")\n",
    "plt.ylabel(r\"Matrix Entropy $\\mathcal{S}(\\mathbf{W}^{e}_{l})$\")\n",
    "#plt.ylim(np.min(fc1_entropies)-0.001, np.max(fc1_entropies)+0.001)\n",
    "plt.tight_layout()\n",
    "plt.legend()\n",
    "\n",
    "plt.savefig(ENTROPY_FILE)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Soft Ranks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:59.213604Z",
     "start_time": "2018-09-14T20:33:47.952459Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7687405b9e69453ab79f887e29725936",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=45), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "fc1_ranks, fc2_ranks = [], []\n",
    "\n",
    "for epoch in tqdm(range(NUM_BASELINE_EPOCHS)):\n",
    "    model.load_weights(BASELINE_per_epoch_weightfile.format(epoch))   \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": 50,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:59.574455Z",
     "start_time": "2018-09-14T20:33:59.215729Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(range(0,len(fc1_ranks)), fc1_ranks, marker='.', color='blue', label='FC1')\n",
    "plt.scatter(range(0,len(fc2_ranks)), fc2_ranks, marker='.', color='teal', label='FC1')\n",
    "#plt.ylim([0.905,0.920])\n",
    "plt.title(r'MLP3:  Stable Rank $\\mathcal{R}(\\mathbf{W}^{e}_{l})$ vs Epoch (e)')\n",
    "plt.xlabel(\"Epoch (e)\")\n",
    "plt.ylabel(r\"Stable Rank $\\mathcal{R}(\\mathbf{W}^{e}_{l})$\")\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(\"img/mlp3-ranks\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Read Baseline Training accuracies from output file\n",
    "\n",
    "Notice: there is no epoch=0 training accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:59.853635Z",
     "start_time": "2018-09-14T20:33:59.576459Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "44 44\n"
     ]
    }
   ],
   "source": [
    "bsln_train_accs =  RMT_Util.read_training_accuracies(BASELINE_outfile)\n",
    "print(len(bsln_train_accs),NUM_BASELINE_EPOCHS-1)\n",
    "assert(len(bsln_train_accs)==NUM_BASELINE_EPOCHS-1)"
   ]
  },
  {
   "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": 52,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:33:59.920226Z",
     "start_time": "2018-09-14T20:33:59.857954Z"
    }
   },
   "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": 53,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:34:03.495773Z",
     "start_time": "2018-09-14T20:33:59.923000Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading test images...\n"
     ]
    }
   ],
   "source": [
    "test_images, test_labels = RMT_Util.read_test_images(datadir=DataDIR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:34:42.604028Z",
     "start_time": "2018-09-14T20:34:03.497238Z"
    }
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1bfe070bb31340a09c9a7d001f1c2a1f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, max=45), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10000/10000 [==============================] - 0s 24us/step\n",
      "10000/10000 [==============================] - 0s 22us/step\n",
      "10000/10000 [==============================] - 0s 22us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 22us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 24us/step\n",
      "10000/10000 [==============================] - 0s 24us/step\n",
      "10000/10000 [==============================] - 0s 25us/step\n",
      "10000/10000 [==============================] - 0s 25us/step\n",
      "10000/10000 [==============================] - 0s 25us/step\n",
      "10000/10000 [==============================] - 0s 26us/step\n",
      "10000/10000 [==============================] - 0s 24us/step\n",
      "10000/10000 [==============================] - 0s 24us/step\n",
      "10000/10000 [==============================] - 0s 25us/step\n",
      "10000/10000 [==============================] - 0s 24us/step\n",
      "10000/10000 [==============================] - 0s 25us/step\n",
      "10000/10000 [==============================] - 0s 27us/step\n",
      "10000/10000 [==============================] - 0s 26us/step\n",
      "10000/10000 [==============================] - 0s 25us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 24us/step\n",
      "10000/10000 [==============================] - 0s 26us/step\n",
      "10000/10000 [==============================] - 0s 24us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 23us/step\n",
      "10000/10000 [==============================] - 0s 27us/step\n",
      "10000/10000 [==============================] - 0s 28us/step\n",
      "10000/10000 [==============================] - 0s 27us/step\n",
      "10000/10000 [==============================] - 0s 31us/step\n",
      "10000/10000 [==============================] - 0s 27us/step\n",
      "10000/10000 [==============================] - 0s 26us/step\n",
      "10000/10000 [==============================] - 0s 25us/step\n",
      "\n"
     ]
    }
   ],
   "source": [
    "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(test_images, test_labels, batch_size=256)\n",
    "    bsln_test_accs.append(acc[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  MLP3 is pretty bad, and easily overtrained</font>\n",
    "\n",
    "### We see how the phase transition corresponds with overtraining on the training data \n",
    "\n",
    "#### But random would be %10, so it is not totally bananas"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:34:42.758935Z",
     "start_time": "2018-09-14T20:34:42.606572Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = range(1,NUM_BASELINE_EPOCHS)\n",
    "\n",
    "#plt.scatter(x, w2_entropies[1:], marker='.', color='blue')\n",
    "#plt.scatter(x, w3_entropies[1:], marker='.', color='green')\n",
    "plt.scatter(x, bsln_train_accs,  marker='.',color='blue', label=\"train\" )\n",
    "plt.scatter(x, bsln_test_accs[1:],  marker='.',color='green', label=\"test\")\n",
    "\n",
    "plt.title('MLP3: Training and Test Accuracies vs Epoch')\n",
    "plt.xlabel(\"Epoch (e)\")\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-04-11T04:12:28.360095Z",
     "start_time": "2018-04-11T04:12:28.357305Z"
    }
   },
   "source": [
    "## Create 9 ESD plots, 3x3 grid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:34:42.765608Z",
     "start_time": "2018-09-14T20:34:42.761473Z"
    }
   },
   "outputs": [],
   "source": [
    "matplotlib.rcParams.update({'font.size': 10})\n",
    "plt.rcParams['figure.figsize'] = [10,10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:34:45.615902Z",
     "start_time": "2018-09-14T20:34:42.768243Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 9 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "LAYER = 2\n",
    "f, axes = plt.subplots(3,3)\n",
    "\n",
    "\n",
    "\n",
    "for idx, epoch in enumerate(np.arange(0,NUM_BASELINE_EPOCHS,5)):\n",
    "    weightfile = BASELINE_per_epoch_weightfile.format(epoch)\n",
    "    eigenvalues = RMT_Util.get_eigenvalues(model, weightfile, layer=LAYER)\n",
    "\n",
    "    iay = idx % 3\n",
    "    iax = int(idx / 3)\n",
    "    #print(idx, iax, iay, modelfile)\n",
    "    subplot = axes[iax,iay]\n",
    "    subplot.hist(eigenvalues, bins=20, alpha=0.25, color='blue', density=True);\n",
    "    subplot.set_title(\"Epoch {}\".format(epoch));\n",
    "\n",
    "    \n",
    "f.subplots_adjust(hspace=0.3)\n",
    "plt.show()\n",
    "\n",
    "#plt.savefig('img/MLP_W2_ESDs.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:34:47.543569Z",
     "start_time": "2018-09-14T20:34:45.618150Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 9 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "LAYER = 3\n",
    "f, axes = plt.subplots(3,3)\n",
    "\n",
    "for idx, epoch in enumerate(np.arange(0,NUM_BASELINE_EPOCHS,5)):\n",
    "    weightfile = BASELINE_per_epoch_weightfile.format(epoch)\n",
    "    eigenvalues = RMT_Util.get_eigenvalues(model, weightfile, layer=LAYER)\n",
    "\n",
    "    iay = idx % 3\n",
    "    iax = int(idx / 3)\n",
    "    #print(idx, iax, iay, modelfile)\n",
    "    subplot = axes[iax,iay]\n",
    "    subplot.hist(eigenvalues, bins=20, alpha=0.25, color='blue', density=True);\n",
    "    subplot.set_title(\"Epoch {}\".format(epoch));\n",
    "\n",
    "    \n",
    "f.subplots_adjust(hspace=0.3)\n",
    "plt.show()\n",
    "\n",
    "#fig = plt.figure()\n",
    "#fig.savefig('MLP_W2_ESDs.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:34:47.548751Z",
     "start_time": "2018-09-14T20:34:47.545434Z"
    }
   },
   "outputs": [],
   "source": [
    "matplotlib.rcParams.update({'font.size': 16})\n",
    "plt.rcParams['figure.figsize'] = [6,5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Fit Marchenko Pastur to ESD\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:37:57.079414Z",
     "start_time": "2018-09-14T20:37:56.177946Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "LAYER=LAYER_FC1\n",
    "EPOCH=0\n",
    "weightfile = BASELINE_per_epoch_weightfile.format(EPOCH)\n",
    "eigenvalues0 = RMT_Util.get_eigenvalues(model, weightfile, layer=LAYER)\n",
    "singular_values0 = np.sqrt(eigenvalues0)\n",
    "\n",
    "weightfile = BASELINE_final_weightfile\n",
    "eigenvalues = RMT_Util.get_eigenvalues(model, weightfile, layer=LAYER)\n",
    "singular_values = np.sqrt(eigenvalues)\n",
    "\n",
    "plt.hist(eigenvalues0,bins=50, density=True, color='red', alpha=0.5, label=\"initial\")\n",
    "plt.hist(eigenvalues,bins=50, density=True, color='blue', alpha=0.5, label=\"final\")\n",
    "plt.title(r\"MLP3 FC1: Empirical Spectral Density $\\rho_{emp}(\\lambda)$\")\n",
    "plt.ylabel(r\"$\\rho_{emp}(\\lambda)$\")\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(FINAL_ESD_FC1_FILE)\n",
    "\n",
    "plt.show()\n",
    "\n",
    "plt.hist(singular_values0,bins=50, density=True, color='red', alpha=0.5, label=\"initial\")\n",
    "plt.hist(singular_values,bins=50, density=True, color='blue', alpha=0.5, label=\"final\")\n",
    "plt.title(r\"MLP3 FC1: Density, Singular Values $\\rho_{emp}(\\nu)$\")\n",
    "plt.ylabel(r\"$\\rho_{emp}(\\nu)$\")\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(FINAL_SEMICIRCLE_FC1_FILE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:41:29.254751Z",
     "start_time": "2018-09-14T20:41:28.042715Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "LAYER=LAYER_FC2\n",
    "EPOCH=0\n",
    "weightfile = BASELINE_per_epoch_weightfile.format(EPOCH)\n",
    "eigenvalues0 = RMT_Util.get_eigenvalues(model, weightfile, layer=LAYER)\n",
    "singular_values0 = np.sqrt(eigenvalues0)\n",
    "\n",
    "weightfile = BASELINE_final_weightfile\n",
    "eigenvalues = RMT_Util.get_eigenvalues(model, weightfile, layer=LAYER)\n",
    "singular_values = np.sqrt(eigenvalues)\n",
    "\n",
    "plt.hist(eigenvalues0,bins=50, density=True, color='red', alpha=0.5, label=\"initial\")\n",
    "plt.hist(eigenvalues,bins=50, density=True, color='blue', alpha=0.5, label=\"final\")\n",
    "plt.title(r\"MLP3 FC2: $\\rho_{emp}(\\lambda)$ Eigenvalues (ESD)\")\n",
    "plt.ylabel(r\"$\\rho_{emp}(\\lambda)$\")\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(FINAL_ESD_FC2_FILE)\n",
    "\n",
    "plt.show()\n",
    "\n",
    "plt.hist(singular_values0,bins=50, density=True, color='red', alpha=0.5, label=\"initial\")\n",
    "plt.hist(singular_values,bins=50, density=True, color='blue', alpha=0.5, label=\"final\")\n",
    "plt.title(r\"MLP3 FC2: $\\rho_{emp}(\\nu)$ Singular Values\")\n",
    "plt.ylabel(r\"$\\rho_{emp}(\\nu)$\")\n",
    "plt.legend()\n",
    "plt.savefig(FINAL_SEMICIRCLE_FC2_FILE)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Scree Plots\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:38:52.250397Z",
     "start_time": "2018-09-14T20:38:51.474130Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "RMT_Util.scree_plot(model, BASELINE_init_weighfile, layer=2, color='red', label=\"FC1\")\n",
    "RMT_Util.scree_plot(model, BASELINE_init_weighfile, layer=3, color='orange', label=\"FC2\")\n",
    "plt.title(r\"MLP3: Scree Plots, initial $\\mathbf{W}_{l}^{0}$\")\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(INIT_SCREE_FILE)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-09-14T20:40:26.079849Z",
     "start_time": "2018-09-14T20:40:25.350985Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "RMT_Util.scree_plot(model, BASELINE_final_weightfile, layer=2, color='blue', label=\"FC1\")\n",
    "RMT_Util.scree_plot(model, BASELINE_final_weightfile, layer=3, color='teal', label=\"FC2\")\n",
    "plt.title(r\"MLP3:  Scree Plots, final $\\mathbf{W}_{f}^{0}$\")\n",
    "plt.legend()\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(FINAL_SCREE_FILE)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": {
    "096cc1f5c1254a84b1885a0669a08db1": {
     "views": [
      {
       "cell_index": 17
      }
     ]
    },
    "aa1b8ea46fb940c7ae68ac77f3d8403c": {
     "views": [
      {
       "cell_index": 13
      }
     ]
    },
    "bf480dfed8274f159e11de14bdb2dc2a": {
     "views": [
      {
       "cell_index": 24
      }
     ]
    }
   },
   "version": "1.2.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}