RNN-3-Non-Ternary.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Input, LSTM, Dense, Dropout, GRU\n",
    "import numpy as np\n",
    "import csv\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import r2_score\n",
    "from keras import optimizers\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "nteams = 20\n",
    "nmatches = (nteams-1)*2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "batch_size = 64  # Batch size for training.\n",
    "epochs = 100  # Number of epochs to train for.\n",
    "latent_dim = 256  # Latent dimensionality of the encoding space.\n",
    "num_samples = 10000  # Number of samples to train on.\n",
    "# Path to the data txt file on disk.\n",
    "data_path = 'fra-eng/fra.txt'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Data Pre-processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def season_team_map(season_dir):\n",
    "    readme = season_dir + '/README.md'\n",
    "    f = open(readme, 'r')\n",
    "    \n",
    "    s = \"  \"\n",
    "    \n",
    "    # skip to teams\n",
    "    while(s[1] != '1'):\n",
    "        s = f.readline()\n",
    "        if(len(s)<=1):\n",
    "            s = \"  \"\n",
    "\n",
    "    team_map = {}\n",
    "\n",
    "    for i in range(20):\n",
    "        team_map[s[4:].split('  ')[0]] = i\n",
    "        s = f.readline()\n",
    "\n",
    "    f.close()\n",
    "    \n",
    "    return team_map"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def season_team_map_div_2(season_dir):\n",
    "    readme = season_dir + '/README.md'\n",
    "    f = open(readme, 'r')\n",
    "    \n",
    "    s = \"  \"\n",
    "    \n",
    "    # skip to teams\n",
    "    while(s[1] != '1'):\n",
    "        s = f.readline()\n",
    "        if(len(s)<=1):\n",
    "            s = \"  \"\n",
    "            \n",
    "    s = f.readline()\n",
    "    if(len(s)<=1):\n",
    "        s = \"  \"    \n",
    "        \n",
    "     # skip to div 2 teams\n",
    "    while(s[:3] != ' 1.'):\n",
    "        s = f.readline()\n",
    "        if(len(s)<=1):\n",
    "            s = \"  \"\n",
    "\n",
    "    team_map = {}\n",
    "\n",
    "    for i in range(20):\n",
    "        team_map[s[4:].split('  ')[0]] = i\n",
    "        s = f.readline()\n",
    "\n",
    "    f.close()\n",
    "    \n",
    "    return team_map"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def season_game_mat(season_dir, team_map, div=1):\n",
    "    game_mat = np.zeros((nteams, nmatches), dtype=int)\n",
    "    team_index = np.zeros(nteams, dtype=int)\n",
    "\n",
    "    csvfile = ''\n",
    "    \n",
    "    for entry in os.scandir(season_dir):\n",
    "        if(div==1):\n",
    "            if '/1-' in entry.path:\n",
    "                csvfile = entry.path\n",
    "        elif(div==2):\n",
    "            if '/2-' in entry.path:\n",
    "                csvfile = entry.path\n",
    "            \n",
    "    with open(csvfile) as gf:\n",
    "        reader = csv.reader(gf)\n",
    "        header = next(reader)\n",
    "        if(header[0] == 'Round'):\n",
    "            new_format = 1\n",
    "        else:\n",
    "            new_format = 0;\n",
    "        for row in reader:\n",
    "            if(new_format):\n",
    "                team1 = team_map[row[2].split(' (')[0]]\n",
    "                team2 = team_map[row[5].split(' (')[0]]\n",
    "\n",
    "                score = row[3].split('-')\n",
    "            else:\n",
    "                team1 = team_map[row[1]]\n",
    "                team2 = team_map[row[2]]\n",
    "\n",
    "                score = row[3].split('-')\n",
    "            # 1 for win, 0 for draw, -1 for loss\n",
    "\n",
    "            game_mat[team1, team_index[team1]] = int(score[0]) - int(score[1])\n",
    "            game_mat[team2, team_index[team2]] = int(score[1]) - int(score[0])\n",
    "\n",
    "            team_index[team1] += 1\n",
    "            team_index[team2] += 1\n",
    "    \n",
    "    return np.transpose(game_mat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "country_dirs = []\n",
    "season_dirs = []\n",
    "\n",
    "nseasons = 0\n",
    "team_maps = []\n",
    "game_mats = []\n",
    "\n",
    "for entry in os.scandir('./data'):\n",
    "    if entry.is_dir():\n",
    "        country_dirs.append(entry.path)\n",
    "\n",
    "for cdir in country_dirs:\n",
    "    for entry in os.scandir(cdir):\n",
    "        if entry.is_dir():\n",
    "            season_dir = entry.path\n",
    "            team_maps.append(season_team_map(season_dir))\n",
    "            game_mats.append(season_game_mat(season_dir, team_maps[nseasons]))\n",
    "            nseasons += 1\n",
    "            if(cdir=='./data/fr-france-master' and int(entry.name.split('-')[0])>=2002):\n",
    "                team_maps.append(season_team_map_div_2(season_dir))\n",
    "                game_mats.append(season_game_mat(season_dir, team_maps[nseasons], div=2))\n",
    "                nseasons += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def eucl_error(outputs, targets):\n",
    "    return np.mean(np.abs(targets - outputs))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def ternarize(x):\n",
    "    if(x > 0.5):\n",
    "        return 1\n",
    "    if(x < -0.5):\n",
    "        return -1\n",
    "    return 0\n",
    "\n",
    "def disc_error(outputs, targets):\n",
    "    ternarize_fn = np.vectorize(ternarize)\n",
    "    ternarized_outputs = ternarize_fn(np.round(outputs))\n",
    "    ternarized_targets = ternarize_fn(np.round(targets))\n",
    "    err = 0\n",
    "    n = 0\n",
    "    for i in range(len(outputs)):\n",
    "        for j in range(len(outputs[i])):\n",
    "            if(int(ternarized_outputs[i][j]) != int(ternarized_targets[i][j])):\n",
    "                err += 1\n",
    "            n += 1\n",
    "    return float(err)/n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "ntraining = int(nseasons*0.7)\n",
    "ntesting = nseasons - ntraining\n",
    "\n",
    "training_game_mats = game_mats[:ntraining]\n",
    "testing_game_mats = game_mats[ntesting:]\n",
    "\n",
    "X_training = np.reshape(np.array(training_game_mats), (ntraining*nmatches, nteams))\n",
    "lengths = np.full(ntraining, nmatches)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "88"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nseasons"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "61"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ntraining"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "Xs = []\n",
    "Ys = []\n",
    "for i in range(1, nmatches):\n",
    "    X = []\n",
    "    Y = []\n",
    "    for game_mat in game_mats:\n",
    "        for j in range(nmatches-i):\n",
    "            X.append(np.ndarray.flatten(game_mat[j:j+i]))\n",
    "            Y.append(game_mat[j+i])\n",
    "    \n",
    "    X = np.array(X).reshape(-1,i,nteams)\n",
    "    Y = np.array(Y).reshape(-1,nteams)\n",
    "    Xs.append(X)\n",
    "    Ys.append(Y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_trainings = []\n",
    "Y_trainings = []\n",
    "for i in range(1, nmatches):\n",
    "    X_training = []\n",
    "    Y_training = []\n",
    "    for game_mat in training_game_mats:\n",
    "        for j in range(nmatches-i):\n",
    "            X_training.append(np.ndarray.flatten(game_mat[j:j+i]))\n",
    "            Y_training.append(game_mat[j+i])\n",
    "    \n",
    "    X_training = np.array(X_training).reshape(-1,i,nteams)\n",
    "    Y_training = np.array(Y_training).reshape(-1,nteams)\n",
    "    X_trainings.append(X_training)\n",
    "    Y_trainings.append(Y_training)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X_testings = []\n",
    "Y_testings = []\n",
    "for i in range(1, nmatches):\n",
    "    X_testing = []\n",
    "    Y_testing = []\n",
    "    for game_mat in testing_game_mats:\n",
    "        for j in range(nmatches-i):\n",
    "            X_testing.append(np.ndarray.flatten(game_mat[j:j+i]))\n",
    "            Y_testing.append(game_mat[j+i])\n",
    "            \n",
    "    X_testing = np.array(X_testing).reshape(-1,i,nteams)\n",
    "    Y_testing = np.array(Y_testing).reshape(-1,nteams)\n",
    "    X_testings.append(X_testing)\n",
    "    Y_testings.append(Y_testing)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "layer_sizes = [200]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "rnns = {}\n",
    "for size in layer_sizes:\n",
    "    key = str(size)+'_neuron_layers'\n",
    "    rnns[key] = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "histories = {}\n",
    "for size in layer_sizes:\n",
    "    key = str(size)+'_neuron_layers'\n",
    "    histories[key] = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "epochs = 20\n",
    "max_sequence_length = 10\n",
    "size = 200"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1830/1830 [==============================] - 7s 4ms/step - loss: 2.6990\n",
      "Epoch 2/10\n",
      "1830/1830 [==============================] - 4s 2ms/step - loss: 2.6448\n",
      "Epoch 3/10\n",
      "1830/1830 [==============================] - 4s 2ms/step - loss: 2.6135A: 3s - loss:\n",
      "Epoch 4/10\n",
      "1830/1830 [==============================] - 4s 2ms/step - loss: 2.5735\n",
      "Epoch 5/10\n",
      "1830/1830 [==============================] - 4s 2ms/step - loss: 2.5184\n",
      "Epoch 6/10\n",
      "1830/1830 [==============================] - 4s 2ms/step - loss: 2.4355\n",
      "Epoch 7/10\n",
      "1830/1830 [==============================] - 4s 2ms/step - loss: 2.3279\n",
      "Epoch 8/10\n",
      "1830/1830 [==============================] - 5s 2ms/step - loss: 2.2167\n",
      "Epoch 9/10\n",
      "1830/1830 [==============================] - 4s 2ms/step - loss: 2.0981\n",
      "Epoch 10/10\n",
      "1830/1830 [==============================] - 4s 2ms/step - loss: 1.9921\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(GRU(size, input_shape=(8, nteams), return_sequences=True))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(GRU(size))\n",
    "model.add(Dropout(0.2))\n",
    "#model.add(Dense(nteams, activation='softmax'))\n",
    "model.add(Dense(nteams, activation='tanh'))\n",
    "opt = optimizers.Adam(lr=0.001)\n",
    "model.compile(loss='mean_squared_error', optimizer=opt)\n",
    "history = model.fit(X_trainings[7], np.array(Y_trainings[7]), batch_size=10, epochs=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "output = model.predict(X_trainings[7])\n",
    "tr_acc = 1 - disc_error(output, np.array(Y_trainings[7]))\n",
    "tr_eerr = eucl_error(output, np.array(Y_trainings[7]))\n",
    "\n",
    "output = model.predict(X_testings[7])\n",
    "te_acc = 1 - disc_error(output, np.array(Y_testings[7]))\n",
    "te_eerr = eucl_error(output, np.array(Y_testings[7]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.6188524590163935 1.00598396692 0.5290437158469945 1.18390556904\n"
     ]
    }
   ],
   "source": [
    "print(tr_acc, tr_eerr, te_acc, te_eerr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['loss'])\n"
     ]
    }
   ],
   "source": [
    "print(history.history.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a44b122b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "2257/2257 [==============================] - 2s 758us/step - loss: 2.7499\n",
      "Epoch 2/20\n",
      "2257/2257 [==============================] - 1s 368us/step - loss: 2.6384\n",
      "Epoch 3/20\n",
      "2257/2257 [==============================] - 0s 203us/step - loss: 2.6140\n",
      "Epoch 4/20\n",
      "2257/2257 [==============================] - 1s 255us/step - loss: 2.6019\n",
      "Epoch 5/20\n",
      "2257/2257 [==============================] - 1s 374us/step - loss: 2.5915\n",
      "Epoch 6/20\n",
      "2257/2257 [==============================] - 0s 192us/step - loss: 2.5844\n",
      "Epoch 7/20\n",
      "2257/2257 [==============================] - 0s 215us/step - loss: 2.5798\n",
      "Epoch 8/20\n",
      "2257/2257 [==============================] - 0s 189us/step - loss: 2.5688\n",
      "Epoch 9/20\n",
      "2257/2257 [==============================] - 1s 246us/step - loss: 2.5665\n",
      "Epoch 10/20\n",
      "2257/2257 [==============================] - 0s 195us/step - loss: 2.5572\n",
      "Epoch 11/20\n",
      "2257/2257 [==============================] - 0s 194us/step - loss: 2.5498\n",
      "Epoch 12/20\n",
      "2257/2257 [==============================] - 0s 202us/step - loss: 2.5440\n",
      "Epoch 13/20\n",
      "2257/2257 [==============================] - 0s 200us/step - loss: 2.5319\n",
      "Epoch 14/20\n",
      "2257/2257 [==============================] - 0s 204us/step - loss: 2.5252\n",
      "Epoch 15/20\n",
      "2257/2257 [==============================] - 0s 218us/step - loss: 2.5228\n",
      "Epoch 16/20\n",
      "2257/2257 [==============================] - 0s 207us/step - loss: 2.5140\n",
      "Epoch 17/20\n",
      "2257/2257 [==============================] - 0s 211us/step - loss: 2.5046\n",
      "Epoch 18/20\n",
      "2257/2257 [==============================] - 0s 210us/step - loss: 2.4955\n",
      "Epoch 19/20\n",
      "2257/2257 [==============================] - 0s 209us/step - loss: 2.4912\n",
      "Epoch 20/20\n",
      "2257/2257 [==============================] - 0s 210us/step - loss: 2.4821\n",
      "Epoch 1/20\n",
      "2196/2196 [==============================] - 2s 1ms/step - loss: 2.7109\n",
      "Epoch 2/20\n",
      "2196/2196 [==============================] - 1s 428us/step - loss: 2.6299\n",
      "Epoch 3/20\n",
      "2196/2196 [==============================] - 1s 295us/step - loss: 2.6072\n",
      "Epoch 4/20\n",
      "2196/2196 [==============================] - 1s 335us/step - loss: 2.5895\n",
      "Epoch 5/20\n",
      "2196/2196 [==============================] - 1s 289us/step - loss: 2.5743\n",
      "Epoch 6/20\n",
      "2196/2196 [==============================] - 1s 294us/step - loss: 2.5544 0s - lo\n",
      "Epoch 7/20\n",
      "2196/2196 [==============================] - 1s 291us/step - loss: 2.5354\n",
      "Epoch 8/20\n",
      "2196/2196 [==============================] - 1s 290us/step - loss: 2.5168\n",
      "Epoch 9/20\n",
      "2196/2196 [==============================] - 1s 344us/step - loss: 2.4880\n",
      "Epoch 10/20\n",
      "2196/2196 [==============================] - 1s 262us/step - loss: 2.4570\n",
      "Epoch 11/20\n",
      "2196/2196 [==============================] - 1s 274us/step - loss: 2.4222\n",
      "Epoch 12/20\n",
      "2196/2196 [==============================] - 1s 251us/step - loss: 2.3869\n",
      "Epoch 13/20\n",
      "2196/2196 [==============================] - 1s 259us/step - loss: 2.3550\n",
      "Epoch 14/20\n",
      "2196/2196 [==============================] - 1s 258us/step - loss: 2.3174\n",
      "Epoch 15/20\n",
      "2196/2196 [==============================] - 1s 267us/step - loss: 2.2765 0s - lo\n",
      "Epoch 16/20\n",
      "2196/2196 [==============================] - 1s 273us/step - loss: 2.2419\n",
      "Epoch 17/20\n",
      "2196/2196 [==============================] - 1s 274us/step - loss: 2.2079\n",
      "Epoch 18/20\n",
      "2196/2196 [==============================] - 1s 276us/step - loss: 2.1747\n",
      "Epoch 19/20\n",
      "2196/2196 [==============================] - 1s 276us/step - loss: 2.1349\n",
      "Epoch 20/20\n",
      "2196/2196 [==============================] - 1s 298us/step - loss: 2.1102\n",
      "Epoch 1/20\n",
      "2135/2135 [==============================] - 2s 720us/step - loss: 2.7030\n",
      "Epoch 2/20\n",
      "2135/2135 [==============================] - 1s 328us/step - loss: 2.6266\n",
      "Epoch 3/20\n",
      "2135/2135 [==============================] - 1s 319us/step - loss: 2.6093\n",
      "Epoch 4/20\n",
      "2135/2135 [==============================] - 1s 382us/step - loss: 2.5931\n",
      "Epoch 5/20\n",
      "2135/2135 [==============================] - 1s 424us/step - loss: 2.5747\n",
      "Epoch 6/20\n",
      "2135/2135 [==============================] - 1s 345us/step - loss: 2.5599\n",
      "Epoch 7/20\n",
      "2135/2135 [==============================] - 1s 332us/step - loss: 2.5400\n",
      "Epoch 8/20\n",
      "2135/2135 [==============================] - 1s 352us/step - loss: 2.5126\n",
      "Epoch 9/20\n",
      "2135/2135 [==============================] - 1s 489us/step - loss: 2.4770\n",
      "Epoch 10/20\n",
      "2135/2135 [==============================] - 1s 354us/step - loss: 2.4407\n",
      "Epoch 11/20\n",
      "2135/2135 [==============================] - 1s 366us/step - loss: 2.3976\n",
      "Epoch 12/20\n",
      "2135/2135 [==============================] - 1s 401us/step - loss: 2.3450\n",
      "Epoch 13/20\n",
      "2135/2135 [==============================] - 1s 519us/step - loss: 2.3012\n",
      "Epoch 14/20\n",
      "2135/2135 [==============================] - 1s 455us/step - loss: 2.2518\n",
      "Epoch 15/20\n",
      "2135/2135 [==============================] - 1s 636us/step - loss: 2.2006\n",
      "Epoch 16/20\n",
      "2135/2135 [==============================] - 1s 427us/step - loss: 2.1513\n",
      "Epoch 17/20\n",
      "2135/2135 [==============================] - 1s 356us/step - loss: 2.0990\n",
      "Epoch 18/20\n",
      "2135/2135 [==============================] - 1s 409us/step - loss: 2.0565 0s\n",
      "Epoch 19/20\n",
      "2135/2135 [==============================] - 1s 415us/step - loss: 2.0142\n",
      "Epoch 20/20\n",
      "2135/2135 [==============================] - 1s 517us/step - loss: 1.9717\n",
      "Epoch 1/20\n",
      "2074/2074 [==============================] - 2s 992us/step - loss: 2.6907\n",
      "Epoch 2/20\n",
      "2074/2074 [==============================] - 1s 435us/step - loss: 2.6248\n",
      "Epoch 3/20\n",
      "2074/2074 [==============================] - 1s 409us/step - loss: 2.6081\n",
      "Epoch 4/20\n",
      "2074/2074 [==============================] - 1s 403us/step - loss: 2.5893\n",
      "Epoch 5/20\n",
      "2074/2074 [==============================] - 1s 396us/step - loss: 2.5712\n",
      "Epoch 6/20\n",
      "2074/2074 [==============================] - 1s 589us/step - loss: 2.5524\n",
      "Epoch 7/20\n",
      "2074/2074 [==============================] - 1s 562us/step - loss: 2.5267 0s - loss:\n",
      "Epoch 8/20\n",
      "2074/2074 [==============================] - 1s 623us/step - loss: 2.5000\n",
      "Epoch 9/20\n",
      "2074/2074 [==============================] - 1s 577us/step - loss: 2.4582\n",
      "Epoch 10/20\n",
      "2074/2074 [==============================] - 1s 541us/step - loss: 2.4118\n",
      "Epoch 11/20\n",
      "2074/2074 [==============================] - 1s 494us/step - loss: 2.3728\n",
      "Epoch 12/20\n",
      "2074/2074 [==============================] - 1s 483us/step - loss: 2.3145\n",
      "Epoch 13/20\n",
      "2074/2074 [==============================] - 1s 483us/step - loss: 2.2613\n",
      "Epoch 14/20\n",
      "2074/2074 [==============================] - 1s 392us/step - loss: 2.2010\n",
      "Epoch 15/20\n",
      "2074/2074 [==============================] - 1s 466us/step - loss: 2.1488\n",
      "Epoch 16/20\n",
      "2074/2074 [==============================] - 1s 419us/step - loss: 2.0972\n",
      "Epoch 17/20\n",
      "2074/2074 [==============================] - 1s 380us/step - loss: 2.0464\n",
      "Epoch 18/20\n",
      "2074/2074 [==============================] - 1s 400us/step - loss: 1.9988\n",
      "Epoch 19/20\n",
      "2074/2074 [==============================] - 1s 416us/step - loss: 1.9530\n",
      "Epoch 20/20\n",
      "2074/2074 [==============================] - 1s 392us/step - loss: 1.9113\n",
      "Epoch 1/20\n",
      "2013/2013 [==============================] - 2s 959us/step - loss: 2.6862\n",
      "Epoch 2/20\n",
      "2013/2013 [==============================] - 1s 400us/step - loss: 2.6284\n",
      "Epoch 3/20\n",
      "2013/2013 [==============================] - 1s 399us/step - loss: 2.6066\n",
      "Epoch 4/20\n",
      "2013/2013 [==============================] - 1s 415us/step - loss: 2.5899\n",
      "Epoch 5/20\n",
      "2013/2013 [==============================] - 1s 439us/step - loss: 2.5708\n",
      "Epoch 6/20\n",
      "2013/2013 [==============================] - 1s 445us/step - loss: 2.5506\n",
      "Epoch 7/20\n",
      "2013/2013 [==============================] - 1s 451us/step - loss: 2.5259\n",
      "Epoch 8/20\n",
      "2013/2013 [==============================] - 1s 451us/step - loss: 2.4948\n",
      "Epoch 9/20\n",
      "2013/2013 [==============================] - 1s 484us/step - loss: 2.4542\n",
      "Epoch 10/20\n",
      "2013/2013 [==============================] - 1s 557us/step - loss: 2.4047\n",
      "Epoch 11/20\n",
      "2013/2013 [==============================] - 1s 536us/step - loss: 2.3478\n",
      "Epoch 12/20\n",
      "2013/2013 [==============================] - 1s 447us/step - loss: 2.2905\n",
      "Epoch 13/20\n",
      "2013/2013 [==============================] - 1s 512us/step - loss: 2.2341\n",
      "Epoch 14/20\n",
      "2013/2013 [==============================] - 1s 485us/step - loss: 2.1723\n",
      "Epoch 15/20\n",
      "2013/2013 [==============================] - 1s 465us/step - loss: 2.1165\n",
      "Epoch 16/20\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2013/2013 [==============================] - 1s 451us/step - loss: 2.0560\n",
      "Epoch 17/20\n",
      "2013/2013 [==============================] - 1s 455us/step - loss: 2.0005\n",
      "Epoch 18/20\n",
      "2013/2013 [==============================] - 1s 489us/step - loss: 1.9549\n",
      "Epoch 19/20\n",
      "2013/2013 [==============================] - 1s 534us/step - loss: 1.9045\n",
      "Epoch 20/20\n",
      "2013/2013 [==============================] - 1s 516us/step - loss: 1.8623\n",
      "Epoch 1/20\n",
      "1952/1952 [==============================] - 2s 1ms/step - loss: 2.6897\n",
      "Epoch 2/20\n",
      "1952/1952 [==============================] - 1s 451us/step - loss: 2.6295\n",
      "Epoch 3/20\n",
      "1952/1952 [==============================] - 1s 489us/step - loss: 2.6123\n",
      "Epoch 4/20\n",
      "1952/1952 [==============================] - 1s 567us/step - loss: 2.5904\n",
      "Epoch 5/20\n",
      "1952/1952 [==============================] - 1s 576us/step - loss: 2.5761\n",
      "Epoch 6/20\n",
      "1952/1952 [==============================] - 1s 596us/step - loss: 2.5553\n",
      "Epoch 7/20\n",
      "1952/1952 [==============================] - 1s 513us/step - loss: 2.5271\n",
      "Epoch 8/20\n",
      "1952/1952 [==============================] - 1s 629us/step - loss: 2.4953\n",
      "Epoch 9/20\n",
      "1952/1952 [==============================] - 1s 673us/step - loss: 2.4537\n",
      "Epoch 10/20\n",
      "1952/1952 [==============================] - 1s 583us/step - loss: 2.4075\n",
      "Epoch 11/20\n",
      "1952/1952 [==============================] - 1s 481us/step - loss: 2.3570\n",
      "Epoch 12/20\n",
      "1952/1952 [==============================] - 1s 533us/step - loss: 2.2903\n",
      "Epoch 13/20\n",
      "1952/1952 [==============================] - 1s 653us/step - loss: 2.2327\n",
      "Epoch 14/20\n",
      "1952/1952 [==============================] - 1s 636us/step - loss: 2.1763\n",
      "Epoch 15/20\n",
      "1952/1952 [==============================] - 1s 739us/step - loss: 2.1181\n",
      "Epoch 16/20\n",
      "1952/1952 [==============================] - 1s 488us/step - loss: 2.0624\n",
      "Epoch 17/20\n",
      "1952/1952 [==============================] - 1s 586us/step - loss: 2.0011\n",
      "Epoch 18/20\n",
      "1952/1952 [==============================] - 2s 799us/step - loss: 1.9541\n",
      "Epoch 19/20\n",
      "1952/1952 [==============================] - 2s 1ms/step - loss: 1.9103\n",
      "Epoch 20/20\n",
      "1952/1952 [==============================] - 2s 863us/step - loss: 1.8625\n",
      "Epoch 1/20\n",
      "1891/1891 [==============================] - 3s 2ms/step - loss: 2.7085\n",
      "Epoch 2/20\n",
      "1891/1891 [==============================] - 1s 531us/step - loss: 2.6412\n",
      "Epoch 3/20\n",
      "1891/1891 [==============================] - 2s 898us/step - loss: 2.6192\n",
      "Epoch 4/20\n",
      "1891/1891 [==============================] - 1s 613us/step - loss: 2.5998\n",
      "Epoch 5/20\n",
      "1891/1891 [==============================] - 1s 537us/step - loss: 2.5766\n",
      "Epoch 6/20\n",
      "1891/1891 [==============================] - 1s 521us/step - loss: 2.5584\n",
      "Epoch 7/20\n",
      "1891/1891 [==============================] - 1s 670us/step - loss: 2.5325\n",
      "Epoch 8/20\n",
      "1891/1891 [==============================] - 1s 654us/step - loss: 2.5026\n",
      "Epoch 9/20\n",
      "1891/1891 [==============================] - 1s 546us/step - loss: 2.4681\n",
      "Epoch 10/20\n",
      "1891/1891 [==============================] - 1s 511us/step - loss: 2.4162\n",
      "Epoch 11/20\n",
      "1891/1891 [==============================] - 1s 564us/step - loss: 2.3749\n",
      "Epoch 12/20\n",
      "1891/1891 [==============================] - 1s 569us/step - loss: 2.3172\n",
      "Epoch 13/20\n",
      "1891/1891 [==============================] - 1s 571us/step - loss: 2.2700\n",
      "Epoch 14/20\n",
      "1891/1891 [==============================] - 1s 531us/step - loss: 2.2104\n",
      "Epoch 15/20\n",
      "1891/1891 [==============================] - 1s 506us/step - loss: 2.1588\n",
      "Epoch 16/20\n",
      "1891/1891 [==============================] - 1s 547us/step - loss: 2.1030\n",
      "Epoch 17/20\n",
      "1891/1891 [==============================] - 1s 513us/step - loss: 2.0478\n",
      "Epoch 18/20\n",
      "1891/1891 [==============================] - 1s 527us/step - loss: 2.0071\n",
      "Epoch 19/20\n",
      "1891/1891 [==============================] - 1s 554us/step - loss: 1.9503\n",
      "Epoch 20/20\n",
      "1891/1891 [==============================] - 2s 901us/step - loss: 1.9086 0s - loss: 1.\n",
      "Epoch 1/20\n",
      "1830/1830 [==============================] - 2s 1ms/step - loss: 2.7004\n",
      "Epoch 2/20\n",
      "1830/1830 [==============================] - 1s 629us/step - loss: 2.6300\n",
      "Epoch 3/20\n",
      "1830/1830 [==============================] - 1s 525us/step - loss: 2.6119\n",
      "Epoch 4/20\n",
      "1830/1830 [==============================] - 1s 528us/step - loss: 2.5900\n",
      "Epoch 5/20\n",
      "1830/1830 [==============================] - 1s 576us/step - loss: 2.5702\n",
      "Epoch 6/20\n",
      "1830/1830 [==============================] - 1s 609us/step - loss: 2.5475\n",
      "Epoch 7/20\n",
      "1830/1830 [==============================] - 1s 645us/step - loss: 2.5180\n",
      "Epoch 8/20\n",
      "1830/1830 [==============================] - 1s 672us/step - loss: 2.4800\n",
      "Epoch 9/20\n",
      "1830/1830 [==============================] - 1s 632us/step - loss: 2.4438\n",
      "Epoch 10/20\n",
      "1830/1830 [==============================] - 2s 898us/step - loss: 2.3974\n",
      "Epoch 11/20\n",
      "1830/1830 [==============================] - 1s 771us/step - loss: 2.3336\n",
      "Epoch 12/20\n",
      "1830/1830 [==============================] - 1s 680us/step - loss: 2.2747\n",
      "Epoch 13/20\n",
      "1830/1830 [==============================] - 1s 644us/step - loss: 2.2211\n",
      "Epoch 14/20\n",
      "1830/1830 [==============================] - 1s 747us/step - loss: 2.1636\n",
      "Epoch 15/20\n",
      "1830/1830 [==============================] - 1s 687us/step - loss: 2.0967\n",
      "Epoch 16/20\n",
      "1830/1830 [==============================] - 1s 661us/step - loss: 2.0377\n",
      "Epoch 17/20\n",
      "1830/1830 [==============================] - 1s 635us/step - loss: 1.9984\n",
      "Epoch 18/20\n",
      "1830/1830 [==============================] - 1s 636us/step - loss: 1.9393\n",
      "Epoch 19/20\n",
      "1830/1830 [==============================] - 1s 649us/step - loss: 1.8991\n",
      "Epoch 20/20\n",
      "1830/1830 [==============================] - 2s 859us/step - loss: 1.8564\n",
      "Epoch 1/20\n",
      "1769/1769 [==============================] - 3s 2ms/step - loss: 2.7079\n",
      "Epoch 2/20\n",
      "1769/1769 [==============================] - 2s 915us/step - loss: 2.6389\n",
      "Epoch 3/20\n",
      "1769/1769 [==============================] - 2s 1ms/step - loss: 2.6198\n",
      "Epoch 4/20\n",
      "1769/1769 [==============================] - 1s 792us/step - loss: 2.6002\n",
      "Epoch 5/20\n",
      "1769/1769 [==============================] - 1s 750us/step - loss: 2.5732\n",
      "Epoch 6/20\n",
      "1769/1769 [==============================] - 1s 798us/step - loss: 2.5507\n",
      "Epoch 7/20\n",
      "1769/1769 [==============================] - 1s 805us/step - loss: 2.5178\n",
      "Epoch 8/20\n",
      "1769/1769 [==============================] - 1s 644us/step - loss: 2.4826\n",
      "Epoch 9/20\n",
      "1769/1769 [==============================] - 1s 544us/step - loss: 2.4371\n",
      "Epoch 10/20\n",
      "1769/1769 [==============================] - 1s 648us/step - loss: 2.3776 0s \n",
      "Epoch 11/20\n",
      "1769/1769 [==============================] - 1s 566us/step - loss: 2.3191\n",
      "Epoch 12/20\n",
      "1769/1769 [==============================] - 1s 597us/step - loss: 2.2581\n",
      "Epoch 13/20\n",
      "1769/1769 [==============================] - 1s 616us/step - loss: 2.2016\n",
      "Epoch 14/20\n",
      "1769/1769 [==============================] - 1s 709us/step - loss: 2.1434 0s - loss: \n",
      "Epoch 15/20\n",
      "1769/1769 [==============================] - 1s 702us/step - loss: 2.0791 0s\n",
      "Epoch 16/20\n",
      "1769/1769 [==============================] - 1s 597us/step - loss: 2.0225\n",
      "Epoch 17/20\n",
      "1769/1769 [==============================] - 1s 653us/step - loss: 1.9616 0s \n",
      "Epoch 18/20\n",
      "1769/1769 [==============================] - 1s 684us/step - loss: 1.9247\n",
      "Epoch 19/20\n",
      "1769/1769 [==============================] - 1s 685us/step - loss: 1.8704\n",
      "Epoch 20/20\n",
      "1769/1769 [==============================] - 1s 628us/step - loss: 1.8277\n",
      "Epoch 1/20\n",
      "1708/1708 [==============================] - 3s 2ms/step - loss: 2.6965\n",
      "Epoch 2/20\n",
      "1708/1708 [==============================] - 1s 576us/step - loss: 2.6388\n",
      "Epoch 3/20\n",
      "1708/1708 [==============================] - 1s 553us/step - loss: 2.6161\n",
      "Epoch 4/20\n",
      "1708/1708 [==============================] - 1s 600us/step - loss: 2.5930\n",
      "Epoch 5/20\n",
      "1708/1708 [==============================] - 1s 734us/step - loss: 2.5745\n",
      "Epoch 6/20\n",
      "1708/1708 [==============================] - 2s 1ms/step - loss: 2.5515A: 0s - loss\n",
      "Epoch 7/20\n",
      "1708/1708 [==============================] - 2s 1ms/step - loss: 2.5202\n",
      "Epoch 8/20\n",
      "1708/1708 [==============================] - 2s 1ms/step - loss: 2.4802\n",
      "Epoch 9/20\n",
      "1708/1708 [==============================] - 2s 1ms/step - loss: 2.4452\n",
      "Epoch 10/20\n",
      "1708/1708 [==============================] - 2s 1ms/step - loss: 2.3880\n",
      "Epoch 11/20\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1708/1708 [==============================] - 2s 1ms/step - loss: 2.3273\n",
      "Epoch 12/20\n",
      "1708/1708 [==============================] - 2s 1ms/step - loss: 2.2686\n",
      "Epoch 13/20\n",
      "1708/1708 [==============================] - 2s 1ms/step - loss: 2.2124A: 1s - loss: 2 - ETA: 1s \n",
      "Epoch 14/20\n",
      "1708/1708 [==============================] - 2s 896us/step - loss: 2.1540\n",
      "Epoch 15/20\n",
      "1708/1708 [==============================] - 2s 1ms/step - loss: 2.0873\n",
      "Epoch 16/20\n",
      "1708/1708 [==============================] - 2s 880us/step - loss: 2.0413\n",
      "Epoch 17/20\n",
      "1708/1708 [==============================] - 1s 870us/step - loss: 1.9791\n",
      "Epoch 18/20\n",
      "1708/1708 [==============================] - 2s 953us/step - loss: 1.9378\n",
      "Epoch 19/20\n",
      "1708/1708 [==============================] - 2s 1ms/step - loss: 1.8881\n",
      "Epoch 20/20\n",
      "1708/1708 [==============================] - 1s 810us/step - loss: 1.8524\n"
     ]
    }
   ],
   "source": [
    "for i in range(max_sequence_length):\n",
    "    print(\"\")\n",
    "    print(\"Sequence Length of \" + str(i))\n",
    "    for size in layer_sizes:\n",
    "        key = str(size)+'_neuron_layers'       \n",
    "        # define the LSTM model\n",
    "        model = Sequential()\n",
    "        model.add(LSTM(size, input_shape=(i+1, nteams)))\n",
    "        model.add(Dropout(0.2))\n",
    "        #model.add(Dense(nteams, activation='softmax'))\n",
    "        model.add(Dense(nteams, activation='tanh'))\n",
    "        #model.compile(loss='categorical_crossentropy', optimizer='adam')\n",
    "        model.compile(loss='mean_squared_error', optimizer='adam')\n",
    "        histories[key].append(model.fit(X_trainings[i], np.array(Y_trainings[i]), epochs=epochs))\n",
    "        rnns[key].append(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def score(key):\n",
    "    training_accs = []\n",
    "    training_errs = []\n",
    "\n",
    "    with open('rnn_nt_training_scores_' + key + '.csv', 'w') as f:\n",
    "        writer = csv.writer(f)\n",
    "        writer.writerow(['Games before prediction', 'Accuracy', 'Euclidean Error'])\n",
    "        print(\"training\")\n",
    "        for i in range(max_sequence_length):\n",
    "            print(i)\n",
    "            outputs = rnns[key][i].predict(X_trainings[i])\n",
    "            acc = 1 - disc_error(outputs, np.array(Y_trainings[i]))\n",
    "            eerr = eucl_error(outputs, np.array(Y_trainings[i]))\n",
    "\n",
    "            writer.writerow([i, acc, eerr])\n",
    "            training_accs.append(acc)\n",
    "            training_errs.append(eerr)\n",
    "            \n",
    "    testing_accs = []\n",
    "    testing_errs = []\n",
    "\n",
    "    with open('rnn_nt_testing_scores_' + key + '.csv', 'w') as f:\n",
    "        writer = csv.writer(f)\n",
    "        writer.writerow(['Games before prediction', 'Accuracy', 'Euclidean Error'])\n",
    "        print(\"testing\")\n",
    "        for i in range(max_sequence_length):\n",
    "            print(i)\n",
    "            outputs = rnns[key][i].predict(X_testings[i])\n",
    "            acc = 1 - disc_error(outputs, np.array(Y_testings[i]))\n",
    "            eerr = eucl_error(outputs, np.array(Y_testings[i]))\n",
    "\n",
    "            writer.writerow([i, acc, eerr])\n",
    "            testing_accs.append(acc)\n",
    "            testing_errs.append(eerr)\n",
    "    \n",
    "    training_accs_dict[key] = training_accs\n",
    "    training_errs_dict[key] = training_errs\n",
    "    testing_accs_dict[key] = testing_accs\n",
    "    testing_errs_dict[key] = testing_errs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "training_accs_dict = {}\n",
    "training_errs_dict = {}\n",
    "testing_accs_dict = {}\n",
    "testing_errs_dict = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nXFwAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a37ff8e10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(histories['200_neuron_layers'][7].history['loss'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "200\n",
      "training\n",
      "0\n",
      "1\n",
      "2\n",
      "3\n",
      "4\n",
      "5\n",
      "6\n",
      "7\n",
      "8\n",
      "9\n",
      "testing\n",
      "0\n",
      "1\n",
      "2\n",
      "3\n",
      "4\n",
      "5\n",
      "6\n",
      "7\n",
      "8\n",
      "9\n"
     ]
    }
   ],
   "source": [
    "for size in layer_sizes:\n",
    "    print(size)\n",
    "    key = str(size)+'_neuron_layers'\n",
    "    score(key)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def plot(y, ylabel, title, imname, ylim):\n",
    "    x = range(1,max_sequence_length+1)\n",
    "    yfit = np.poly1d(np.polyfit(x, y, 1))(np.unique(x))\n",
    "    r2 = np.round(r2_score(y, yfit),decimals=2)\n",
    "    mean = np.round(np.mean(y), decimals=2)\n",
    "\n",
    "    plt.scatter(x, y, s = 5)\n",
    "    plt.plot(x, yfit)\n",
    "    plt.axis([0,max_sequence_length+1,0,ylim])\n",
    "    plt.text(1, 0.95, 'R^2: ' + str(r2))\n",
    "    plt.text(1, 0.85, 'Mean: ' + str(mean))\n",
    "    plt.xlabel('Games before prediction')\n",
    "    plt.ylabel(ylabel)\n",
    "    plt.title(title)\n",
    "    plt.savefig('rnnntplots/' + imname + '.png')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a37bd6fd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY8AAAEWCAYAAACe8xtsAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a3942ee80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a2f050b00>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a2dbfd5c0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for size in layer_sizes:\n",
    "    key = str(size)+'_neuron_layers'\n",
    "    title = ' ' + str(size) + ' LSTM Layer'\n",
    "    plot(training_errs_dict[key], 'Euclidian Error', 'RNN Non-Ternary Training Euclidean Error' + title, 'rnn_nt_training_error_plot_'+key, 2)\n",
    "    plot(testing_errs_dict[key], 'Euclidian Error', 'RNN Non-Ternary Testing Euclidean Error' + title, 'rnn_nt_testing_error_plot_'+key, 2)\n",
    "    plot(training_accs_dict[key], 'Accuracy', 'RNN Non-Ternary Training Accuracy' + title, 'rnn_nt_training_accuracy_plot_'+key, 1)\n",
    "    plot(testing_accs_dict[key], 'Accuracy', 'RNN Non-Ternary Testing Accuracy' + title, 'rnn_nt_testing_accuracy_plot_'+key, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}