SenseGenModel_ECGDataset.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\nAuthor: Moustafa Alzantot (malzantot@ucla.edu)\\nAll rights reserved Networked and Embedded Systems Lab (NESL), UCLA.\\nPermission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the \"Software\"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:\\nThe above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.\\nTHE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\\n'"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\"\"\"\n",
    "Author: Moustafa Alzantot (malzantot@ucla.edu)\n",
    "All rights reserved Networked and Embedded Systems Lab (NESL), UCLA.\n",
    "Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the \"Software\"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:\n",
    "The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.\n",
    "THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/malzantot/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"
     ]
    }
   ],
   "source": [
    "import data_utils\n",
    "import model_utils\n",
    "import model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data_utils.load_training_data('ecg')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# To get reasonable outputs, should use something bigger than 1000 !\n",
    "num_epochs = 1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/malzantot/Nesl/sensegen/model.py:137: LSTMCell.__init__ (from tensorflow.python.ops.rnn_cell_impl) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "This class is equivalent as tf.keras.layers.LSTMCell, and will be replaced by that in Tensorflow 2.0.\n",
      "WARNING:tensorflow:From /home/malzantot/Nesl/sensegen/model.py:137: MultiRNNCell.__init__ (from tensorflow.python.ops.rnn_cell_impl) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "This class is equivalent as tf.keras.layers.StackedRNNCells, and will be replaced by that in Tensorflow 2.0.\n",
      "WARNING:tensorflow:From /home/malzantot/Nesl/sensegen/model.py:142: dynamic_rnn (from tensorflow.python.ops.rnn) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `keras.layers.RNN(cell)`, which is equivalent to this API\n",
      "WARNING:tensorflow:From /home/malzantot/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/tensor_array_ops.py:162: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "Tensor(\"mdn_model/add_1:0\", shape=(1280, 72), dtype=float32)\n",
      "Tensor(\"mdn_model/strided_slice:0\", shape=(1280, 24), dtype=float32)\n",
      "Tensor(\"mdn_model/Exp:0\", shape=(1280, 24), dtype=float32)\n",
      "Tensor(\"mdn_model/y:0\", shape=(128, 10, 1), dtype=float32)\n",
      "\n",
      "WARNING: The TensorFlow contrib module will not be included in TensorFlow 2.0.\n",
      "For more information, please see:\n",
      "  * https://github.com/tensorflow/community/blob/master/rfcs/20180907-contrib-sunset.md\n",
      "  * https://github.com/tensorflow/addons\n",
      "If you depend on functionality not listed there, please file an issue.\n",
      "\n",
      "WARNING:tensorflow:From /home/malzantot/Nesl/sensegen/model.py:169: Normal.__init__ (from tensorflow.python.ops.distributions.normal) is deprecated and will be removed after 2019-01-01.\n",
      "Instructions for updating:\n",
      "The TensorFlow Distributions library has moved to TensorFlow Probability (https://github.com/tensorflow/probability). You should update all references to use `tfp.distributions` instead of `tf.distributions`.\n",
      "WARNING:tensorflow:From /home/malzantot/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/distributions/normal.py:160: Distribution.__init__ (from tensorflow.python.ops.distributions.distribution) is deprecated and will be removed after 2019-01-01.\n",
      "Instructions for updating:\n",
      "The TensorFlow Distributions library has moved to TensorFlow Probability (https://github.com/tensorflow/probability). You should update all references to use `tfp.distributions` instead of `tf.distributions`.\n",
      "WARNING:tensorflow:From /home/malzantot/Nesl/sensegen/model.py:171: calling reduce_sum_v1 (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "keep_dims is deprecated, use keepdims instead\n",
      "WARNING:tensorflow:From /home/malzantot/anaconda3/lib/python3.6/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Tensor(\"mdn_model_1/add_1:0\", shape=(1, 72), dtype=float32)\n",
      "Tensor(\"mdn_model_1/strided_slice:0\", shape=(1, 24), dtype=float32)\n",
      "Tensor(\"mdn_model_1/Exp:0\", shape=(1, 24), dtype=float32)\n",
      "0   -0.30225736\n",
      "1   -0.9016295\n",
      "2   -0.9758881\n",
      "3   -1.0264001\n",
      "4   -1.1310802\n",
      "5   -1.1699288\n",
      "6   -1.1880974\n",
      "7   -1.1973692\n",
      "8   -1.2342938\n",
      "9   -1.2528385\n",
      "10   -1.261825\n",
      "11   -1.2780408\n",
      "12   -1.2803981\n",
      "13   -1.2868296\n",
      "14   -1.298619\n",
      "15   -1.3030744\n",
      "16   -1.3237638\n",
      "17   -1.3201647\n",
      "18   -1.3323042\n",
      "19   -1.3246816\n",
      "20   -1.3321375\n",
      "21   -1.3409641\n",
      "22   -1.3422601\n",
      "23   -1.3530954\n",
      "24   -1.3538653\n",
      "25   -1.3550696\n",
      "26   -1.3602246\n",
      "27   -1.3558002\n",
      "28   -1.3615742\n",
      "29   -1.3753655\n",
      "30   -1.3806916\n",
      "31   -1.3713285\n",
      "32   -1.3701558\n",
      "33   -1.3812367\n",
      "34   -1.3835685\n",
      "35   -1.3796666\n",
      "36   -1.375839\n",
      "37   -1.3942122\n",
      "38   -1.388011\n",
      "39   -1.3989283\n",
      "40   -1.3890725\n",
      "41   -1.3946867\n",
      "42   -1.4031006\n",
      "43   -1.4111859\n",
      "44   -1.4109526\n",
      "45   -1.4103806\n",
      "46   -1.4118586\n",
      "47   -1.4216504\n",
      "48   -1.4210384\n",
      "49   -1.4238687\n",
      "50   -1.4236597\n",
      "51   -1.4169451\n",
      "52   -1.4255313\n",
      "53   -1.4312707\n",
      "54   -1.4274695\n",
      "55   -1.4373108\n",
      "56   -1.4419549\n",
      "57   -1.4437859\n",
      "58   -1.426924\n",
      "59   -1.4486983\n",
      "60   -1.444116\n",
      "61   -1.4485886\n",
      "62   -1.446899\n",
      "63   -1.4473901\n",
      "64   -1.4529332\n",
      "65   -1.4553167\n",
      "66   -1.4519411\n",
      "67   -1.4595153\n",
      "68   -1.4596541\n",
      "69   -1.4455043\n",
      "70   -1.4526408\n",
      "71   -1.4584342\n",
      "72   -1.4582187\n",
      "73   -1.4604298\n",
      "74   -1.4592882\n",
      "75   -1.4596982\n",
      "76   -1.4626045\n",
      "77   -1.462644\n",
      "78   -1.4595338\n",
      "79   -1.4653779\n",
      "80   -1.4624964\n",
      "81   -1.4660225\n",
      "82   -1.463485\n",
      "83   -1.4689147\n",
      "84   -1.4696825\n",
      "85   -1.4690361\n",
      "86   -1.4657669\n",
      "87   -1.4694601\n",
      "88   -1.4666291\n",
      "89   -1.4715672\n",
      "90   -1.4681233\n",
      "91   -1.476541\n",
      "92   -1.471577\n",
      "93   -1.4772997\n",
      "94   -1.4752549\n",
      "95   -1.4717618\n",
      "96   -1.4752661\n",
      "97   -1.476192\n",
      "98   -1.4785197\n",
      "99   -1.4775305\n",
      "100   -1.4787151\n",
      "101   -1.4812932\n",
      "102   -1.4807873\n",
      "103   -1.4828718\n",
      "104   -1.4780173\n",
      "105   -1.484727\n",
      "106   -1.4839545\n",
      "107   -1.4771155\n",
      "108   -1.4815919\n",
      "109   -1.4721901\n",
      "110   -1.4843374\n",
      "111   -1.4888343\n",
      "112   -1.490261\n",
      "113   -1.4852774\n",
      "114   -1.4918458\n",
      "115   -1.4888762\n",
      "116   -1.4893463\n",
      "117   -1.4919678\n",
      "118   -1.4854027\n",
      "119   -1.4868699\n",
      "120   -1.4812721\n",
      "121   -1.4877647\n",
      "122   -1.4881933\n",
      "123   -1.4769492\n",
      "124   -1.4821999\n",
      "125   -1.492864\n",
      "126   -1.4875646\n",
      "127   -1.4948552\n",
      "128   -1.490339\n",
      "129   -1.4845905\n",
      "130   -1.4950291\n",
      "131   -1.4909778\n",
      "132   -1.4815927\n",
      "133   -1.4976249\n",
      "134   -1.4978878\n",
      "135   -1.4969481\n",
      "136   -1.4952134\n",
      "137   -1.4937725\n",
      "138   -1.4965197\n",
      "139   -1.4969244\n",
      "140   -1.5009148\n",
      "141   -1.5002909\n",
      "142   -1.4986441\n",
      "143   -1.5044045\n",
      "144   -1.4968405\n",
      "145   -1.5018941\n",
      "146   -1.500159\n",
      "147   -1.5066503\n",
      "148   -1.5098878\n",
      "149   -1.504028\n",
      "150   -1.5044353\n",
      "151   -1.5056092\n",
      "152   -1.5030725\n",
      "153   -1.5084218\n",
      "154   -1.5047723\n",
      "155   -1.5048201\n",
      "156   -1.5057611\n",
      "157   -1.5053183\n",
      "158   -1.5057329\n",
      "159   -1.5080662\n",
      "160   -1.4997542\n",
      "161   -1.508608\n",
      "162   -1.5057608\n",
      "163   -1.504062\n",
      "164   -1.5117438\n",
      "165   -1.5055296\n",
      "166   -1.5049939\n",
      "167   -1.5120149\n",
      "168   -1.509856\n",
      "169   -1.5078015\n",
      "170   -1.5042849\n",
      "171   -1.5077991\n",
      "172   -1.5129814\n",
      "173   -1.4999243\n",
      "174   -1.5097889\n",
      "175   -1.5127192\n",
      "176   -1.5133644\n",
      "177   -1.5117797\n",
      "178   -1.5143368\n",
      "179   -1.5065318\n",
      "180   -1.5122741\n",
      "181   -1.5068969\n",
      "182   -1.5117996\n",
      "183   -1.5132699\n",
      "184   -1.5110726\n",
      "185   -1.5098923\n",
      "186   -1.5114975\n",
      "187   -1.5112145\n",
      "188   -1.5109147\n",
      "189   -1.508002\n",
      "190   -1.5121273\n",
      "191   -1.5068486\n",
      "192   -1.5096141\n",
      "193   -1.5102317\n",
      "194   -1.5046748\n",
      "195   -1.504132\n",
      "196   -1.5102348\n",
      "197   -1.513514\n",
      "198   -1.5076952\n",
      "199   -1.5075676\n",
      "200   -1.5114856\n",
      "201   -1.5099221\n",
      "202   -1.5101441\n",
      "203   -1.5110042\n",
      "204   -1.5105345\n",
      "205   -1.5154942\n",
      "206   -1.5131586\n",
      "207   -1.5103248\n",
      "208   -1.5105921\n",
      "209   -1.5092493\n",
      "210   -1.5116601\n",
      "211   -1.5108223\n",
      "212   -1.5150684\n",
      "213   -1.5154699\n",
      "214   -1.5083084\n",
      "215   -1.5118449\n",
      "216   -1.5129875\n",
      "217   -1.5133694\n",
      "218   -1.509227\n",
      "219   -1.5124044\n",
      "220   -1.5121838\n",
      "221   -1.5136706\n",
      "222   -1.5094274\n",
      "223   -1.5158398\n",
      "224   -1.5150388\n",
      "225   -1.5133282\n",
      "226   -1.5142815\n",
      "227   -1.510927\n",
      "228   -1.5120956\n",
      "229   -1.5140282\n",
      "230   -1.5202981\n",
      "231   -1.5185134\n",
      "232   -1.5123768\n",
      "233   -1.5117375\n",
      "234   -1.515057\n",
      "235   -1.5128115\n",
      "236   -1.5160122\n",
      "237   -1.5133728\n",
      "238   -1.5168695\n",
      "239   -1.5138264\n",
      "240   -1.5165799\n",
      "241   -1.5202827\n",
      "242   -1.5123134\n",
      "243   -1.5141369\n",
      "244   -1.5115983\n",
      "245   -1.5123922\n",
      "246   -1.5174929\n",
      "247   -1.5149935\n",
      "248   -1.5130029\n",
      "249   -1.5083704\n",
      "250   -1.5008992\n",
      "251   -1.5143211\n",
      "252   -1.5237432\n",
      "253   -1.5186696\n",
      "254   -1.5202729\n",
      "255   -1.5170041\n",
      "256   -1.5166199\n",
      "257   -1.5178015\n",
      "258   -1.5227232\n",
      "259   -1.5100086\n",
      "260   -1.516942\n",
      "261   -1.5221556\n",
      "262   -1.5125202\n",
      "263   -1.5166626\n",
      "264   -1.5207931\n",
      "265   -1.5246634\n",
      "266   -1.5295227\n",
      "267   -1.5275145\n",
      "268   -1.5211431\n",
      "269   -1.5234003\n",
      "270   -1.5196128\n",
      "271   -1.526211\n",
      "272   -1.5201617\n",
      "273   -1.5161494\n",
      "274   -1.5195906\n",
      "275   -1.527176\n",
      "276   -1.5247585\n",
      "277   -1.5190821\n",
      "278   -1.525553\n",
      "279   -1.523774\n",
      "280   -1.5268039\n",
      "281   -1.5256324\n",
      "282   -1.5187738\n",
      "283   -1.5274687\n",
      "284   -1.5173954\n",
      "285   -1.5101644\n",
      "286   -1.5179585\n",
      "287   -1.5235218\n",
      "288   -1.5261014\n",
      "289   -1.5290205\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "290   -1.5276003\n",
      "291   -1.526595\n",
      "292   -1.5217355\n",
      "293   -1.5268391\n",
      "294   -1.5254116\n",
      "295   -1.5270076\n",
      "296   -1.5166708\n",
      "297   -1.5180691\n",
      "298   -1.5237727\n",
      "299   -1.5181568\n",
      "300   -1.5242374\n",
      "301   -1.5289862\n",
      "302   -1.528128\n",
      "303   -1.5272272\n",
      "304   -1.524831\n",
      "305   -1.5213408\n",
      "306   -1.5192943\n",
      "307   -1.5211384\n",
      "308   -1.530019\n",
      "309   -1.5263447\n",
      "310   -1.5278682\n",
      "311   -1.5307882\n",
      "312   -1.5252777\n",
      "313   -1.5266764\n",
      "314   -1.5286514\n",
      "315   -1.5309445\n",
      "316   -1.5297023\n",
      "317   -1.5293146\n",
      "318   -1.5215061\n",
      "319   -1.5273637\n",
      "320   -1.5281324\n",
      "321   -1.530785\n",
      "322   -1.5276033\n",
      "323   -1.5312023\n",
      "324   -1.5239483\n",
      "325   -1.5278238\n",
      "326   -1.5312405\n",
      "327   -1.5315282\n",
      "328   -1.5313369\n",
      "329   -1.5274658\n",
      "330   -1.531447\n",
      "331   -1.5272026\n",
      "332   -1.5241928\n",
      "333   -1.530846\n",
      "334   -1.5275483\n",
      "335   -1.5327772\n",
      "336   -1.5306683\n",
      "337   -1.5268233\n",
      "338   -1.526803\n",
      "339   -1.5323822\n",
      "340   -1.5249627\n",
      "341   -1.5312594\n",
      "342   -1.5224974\n",
      "343   -1.5228763\n",
      "344   -1.5323776\n",
      "345   -1.5249143\n",
      "346   -1.5284789\n",
      "347   -1.5270176\n",
      "348   -1.5300974\n",
      "349   -1.5293187\n",
      "350   -1.5279602\n",
      "351   -1.5197133\n",
      "352   -1.5292267\n",
      "353   -1.5182769\n",
      "354   -1.5220273\n",
      "355   -1.5293491\n",
      "356   -1.5247715\n",
      "357   -1.5237035\n",
      "358   -1.5304126\n",
      "359   -1.5284162\n",
      "360   -1.5320547\n",
      "361   -1.519982\n",
      "362   -1.5288659\n",
      "363   -1.5269455\n",
      "364   -1.5200175\n",
      "365   -1.525511\n",
      "366   -1.525453\n",
      "367   -1.5255369\n",
      "368   -1.5223013\n",
      "369   -1.531625\n",
      "370   -1.5324117\n",
      "371   -1.5296425\n",
      "372   -1.5276752\n",
      "373   -1.5329815\n",
      "374   -1.5297412\n",
      "375   -1.5290095\n",
      "376   -1.5315715\n",
      "377   -1.5318837\n",
      "378   -1.5297729\n",
      "379   -1.5215596\n",
      "380   -1.5360425\n",
      "381   -1.5321892\n",
      "382   -1.532932\n",
      "383   -1.5289149\n",
      "384   -1.5318658\n",
      "385   -1.5302466\n",
      "386   -1.5342908\n",
      "387   -1.5274019\n",
      "388   -1.5325809\n",
      "389   -1.5345662\n",
      "390   -1.5339288\n",
      "391   -1.5312002\n",
      "392   -1.5296836\n",
      "393   -1.5252416\n",
      "394   -1.5309366\n",
      "395   -1.5332506\n",
      "396   -1.5334619\n",
      "397   -1.5359544\n",
      "398   -1.5284281\n",
      "399   -1.5376202\n",
      "400   -1.5310094\n",
      "401   -1.5351522\n",
      "402   -1.5362556\n",
      "403   -1.5293115\n",
      "404   -1.5338084\n",
      "405   -1.5322225\n",
      "406   -1.5319463\n",
      "407   -1.5354942\n",
      "408   -1.5331061\n",
      "409   -1.5368834\n",
      "410   -1.5373237\n",
      "411   -1.5329795\n",
      "412   -1.5356042\n",
      "413   -1.5374467\n",
      "414   -1.5344105\n",
      "415   -1.5393586\n",
      "416   -1.5312672\n",
      "417   -1.5361015\n",
      "418   -1.5382856\n",
      "419   -1.5382527\n",
      "420   -1.5330033\n",
      "421   -1.5371863\n",
      "422   -1.5346358\n",
      "423   -1.531743\n",
      "424   -1.53284\n",
      "425   -1.5348405\n"
     ]
    }
   ],
   "source": [
    "model_utils.reset_session_and_model()\n",
    "with tf.Session() as sess:\n",
    "    train_config = model.ModelConfig()\n",
    "    test_config = model.ModelConfig()\n",
    "    train_config.learning_rate = 0.003\n",
    "    train_config.num_layers = 1\n",
    "    train_config.batch_size = 128\n",
    "    test_config.num_layers = 1\n",
    "    test_config.batch_size = 1\n",
    "    test_config.num_steps = 1\n",
    "    loader = data_utils.DataLoader(data=data,batch_size=train_config.batch_size, num_steps=train_config.num_steps)\n",
    "    train_model = model.MDNModel(train_config, True)\n",
    "    test_model = model.MDNModel(test_config, False)\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    saver = tf.train.Saver()\n",
    "    for idx in range(num_epochs):\n",
    "        epoch_loss = train_model.train_for_epoch(sess, loader)\n",
    "        print(idx, ' ', epoch_loss)\n",
    "        if (idx+1) % 100 == 0:\n",
    "            saver.save(sess, './models/ecg_mdnmodel.ckpt', global_step=idx)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## Sampling from a trained model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensor(\"mdn_model/add_1:0\", shape=(1, 72), dtype=float32)\n",
      "Tensor(\"mdn_model/strided_slice:0\", shape=(1, 24), dtype=float32)\n",
      "Tensor(\"mdn_model/Exp:0\", shape=(1, 24), dtype=float32)\n",
      "Tensor(\"mdn_model/y:0\", shape=(1, 1, 1), dtype=float32)\n",
      "INFO:tensorflow:Restoring parameters from ./models/ecg_mdnmodel.ckpt-999\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'Fake data')"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x576 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ckpt_path = './models/ecg_mdnmodel.ckpt-999'\n",
    "seq_len = 1200\n",
    "model_utils.reset_session_and_model()\n",
    "true_data = data[0,:seq_len]\n",
    "with tf.Session() as sess:\n",
    "    test_config = model.ModelConfig()\n",
    "    test_config.num_layers = 1\n",
    "    test_config.batch_size = 1\n",
    "    test_config.num_steps = 1\n",
    "    test_model = model.MDNModel(test_config, True)\n",
    "    test_model.is_training = False\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    saver = tf.train.Saver()\n",
    "    saver.restore(sess, ckpt_path)\n",
    "    fake_data = test_model.predict(sess, seq_len)\n",
    "fig, axes = plt.subplots(1,2, figsize=((14,8)))\n",
    "axes[0].plot(true_data)\n",
    "axes[0].set_title('True data')\n",
    "axes[1].plot(fake_data)\n",
    "axes[1].set_title('Fake data')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Note:\n",
    "# I didn't spend time to do hyperparameter selection for the model and improve the results.\n",
    "# Changing the number of Guassian mixtures, number of hidden units , number of epochs, or learning rate should improve the quality of results.\n",
    "# Another useful trick would be to 'quantize' the signal into discrete set of levels and use one-hot-encoding for input processing and cross-entropy loss.\n",
    "# Under such a setting, the model will be similar to the RNNLM architecture trained using maximum likelihood estimate (MLE).\n",
    "# Sometimes the model above makes a mistakce and produce signal that not looking well-shaped, this is mainly due to that MLE-based models\n",
    "# suffer from `exposure bias` which means at generation time the model may produce data different from those seen during training time, and since\n",
    "# the model relies on its own prediction as an input for next step, the error would propagate to future time steps as well.\n",
    "# Further direction of improvement would be pairing the generator model with a discriminator and train the generator by adversarial loss.\n",
    "\n"
   ]
  },
  {
   "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"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}