ANN_RNAP_Seq_Classifier.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Artificial Neural Network that predicts whether RNAP will bind to an arbitrary sequence\n",
    "\n",
    "(Trained on the Ecoli genome)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "input_data_raw =  pd.read_csv('rnap_scanned_over_ecoli_genome.csv')\n",
    "sequences = input_data_raw['seq'].tolist()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Raw Data\n",
    "The seq column below contains positive training examples. Sequences that RNAP will bind to. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>val</th>\n",
       "      <th>seq</th>\n",
       "      <th>left</th>\n",
       "      <th>right</th>\n",
       "      <th>ori</th>\n",
       "      <th>contig</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>23.9083</td>\n",
       "      <td>CTATTACCTGCCCACGGTAAAGGCGTTCTGGTTGCCCTACG</td>\n",
       "      <td>3695629</td>\n",
       "      <td>3695669</td>\n",
       "      <td>+</td>\n",
       "      <td>MG1655.fa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>23.5831</td>\n",
       "      <td>TTTCTGCCCGCGGCCTGGCAATTGATACCGCCTTGCCATCG</td>\n",
       "      <td>1109059</td>\n",
       "      <td>1109099</td>\n",
       "      <td>+</td>\n",
       "      <td>MG1655.fa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>23.4205</td>\n",
       "      <td>TTTTTACACGCCTTGCGTAAAGCGGCACGGCGCGGGGTGCG</td>\n",
       "      <td>822216</td>\n",
       "      <td>822256</td>\n",
       "      <td>-</td>\n",
       "      <td>MG1655.fa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>23.4098</td>\n",
       "      <td>TTGCTGCCTGGCCGATGGTCTATGCCGCTGCGTTCTCCGGC</td>\n",
       "      <td>772478</td>\n",
       "      <td>772518</td>\n",
       "      <td>+</td>\n",
       "      <td>MG1655.fa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>23.3640</td>\n",
       "      <td>CTTTTACACGCAAAATGCCGTCACGTTCACGGGTCTGGGCT</td>\n",
       "      <td>2748531</td>\n",
       "      <td>2748571</td>\n",
       "      <td>+</td>\n",
       "      <td>MG1655.fa</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       val                                        seq     left    right ori  \\\n",
       "0  23.9083  CTATTACCTGCCCACGGTAAAGGCGTTCTGGTTGCCCTACG  3695629  3695669   +   \n",
       "1  23.5831  TTTCTGCCCGCGGCCTGGCAATTGATACCGCCTTGCCATCG  1109059  1109099   +   \n",
       "2  23.4205  TTTTTACACGCCTTGCGTAAAGCGGCACGGCGCGGGGTGCG   822216   822256   -   \n",
       "3  23.4098  TTGCTGCCTGGCCGATGGTCTATGCCGCTGCGTTCTCCGGC   772478   772518   +   \n",
       "4  23.3640  CTTTTACACGCAAAATGCCGTCACGTTCACGGGTCTGGGCT  2748531  2748571   +   \n",
       "\n",
       "      contig  \n",
       "0  MG1655.fa  \n",
       "1  MG1655.fa  \n",
       "2  MG1655.fa  \n",
       "3  MG1655.fa  \n",
       "4  MG1655.fa  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "input_data_raw.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Set up lists that will contain positive sequences and positive test sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#np.array([np.random.choice(['A', 'C', 'G', 'T'], 1, p=np.array([1,1,1,1])/4.0) for i in range(seq_length)]).transpose()\n",
    "np.random.seed(0)\n",
    "random_seq_indices = np.random.randint(9000, size = (1000))\n",
    "test_random_seq_indices = np.random.randint(1000, size = (500))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "pos = []\n",
    "test_pos = []\n",
    "\n",
    "num_sample = 9000\n",
    "test_size = 1000\n",
    "\n",
    "seq_length = 41\n",
    "\n",
    "for index in range(len(sequences[0:num_sample])):\n",
    "    #if(index in random_seq_indices):\n",
    "    #    pos.append(np.array([np.random.choice(['A', 'C', 'G', 'T'], p=np.array([1,1,1,1])/4.0) for i in range(41)]).transpose())\n",
    "    #else:    \n",
    "        pos.append(list(sequences[index]))\n",
    "    \n",
    "for test_index in range(len(sequences[num_sample:num_sample+test_size])):\n",
    "    #if(test_index in test_random_seq_indices):\n",
    "    #    test_pos.append(np.array([np.random.choice(['A', 'C', 'G', 'T'], p=np.array([1,1,1,1])/4.0) for i in range(41)]).transpose())\n",
    "    #else:\n",
    "        test_pos.append(list(sequences[test_index]))    \n",
    "\n",
    "pos = np.array(pos)\n",
    "test_pos = np.array(test_pos)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "... and negative training and test sequences "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 170
    },
    "colab_type": "code",
    "id": "_FyVOov3NoFy",
    "outputId": "77e43c57-5fa1-43f4-b21f-eeb17e06e785"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['TCGTGGAGAAGCCCTAGATGAGCGTCACATGGAGCCATCGC',\n",
       " 'CAAGCAATGGAAAAGTATTTCGGGGCCTCCTCTCTTGAACC',\n",
       " 'TCTATGTAGGTCGCTGCCCGACTCGGGGTTGGCGCCTCAAC',\n",
       " 'ACCCGCGACGCCGAGAGCACGCGGTTGTTCATTTTGGGCCA',\n",
       " 'GTAAAATCTTTGGTTCATCTTAGGTTATCCTCGGAGAATCT',\n",
       " 'TGGTACTAGAGGGCGTTGAGACCTCCGTGCGCAAATATGGC',\n",
       " 'AGAAAAGTTAAAAGAAATCGCTAAGGTGATACTCAAGACCC',\n",
       " 'CTACCTTCGTAATCGTGAGTTTGGAGCGTGTCCAGATCTCT',\n",
       " 'TAGGACTCCGATTGGTTCACAGACTAATTCCTCTGTTCCAA']"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "neg = np.array([np.random.choice(['A', 'C', 'G', 'T'], num_sample, p=np.array([1,1,1,1])/4.0)\n",
    "                for i in range(seq_length)]).transpose()\n",
    "\n",
    "test_neg = np.array([np.random.choice(['A', 'C', 'G', 'T'], test_size, p=np.array([1,1,1,1])/4.0)\n",
    "                for i in range(seq_length)]).transpose()\n",
    "\n",
    "[''.join(x) for x in neg[1:10,:]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(9000, 41)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.shape(pos)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Change training examples to one-hot encoded tensors\n",
    "\n",
    "### and put labels in tensors called 'y' and 'y_test'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 289
    },
    "colab_type": "code",
    "id": "riQRzV6CNp33",
    "outputId": "a27104ff-7917-4987-dccc-7780792beac2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Positive sample matrix shape: (9000, 41)\n",
      "Positive sample tensor shape: (9000, 41, 4)\n",
      "Training set shape: (18000, 41, 4)\n",
      "Training set label shape: (18000,)\n",
      "\n",
      "One-hot encoding looks like:\n",
      " [[0. 1. 0. 0.]\n",
      " [0. 0. 0. 1.]\n",
      " [1. 0. 0. 0.]\n",
      " [0. 0. 0. 1.]\n",
      " [0. 0. 0. 1.]\n",
      " [1. 0. 0. 0.]\n",
      " [0. 1. 0. 0.]\n",
      " [0. 1. 0. 0.]\n",
      " [0. 0. 0. 1.]\n",
      " [0. 0. 1. 0.]]\n"
     ]
    }
   ],
   "source": [
    "pos_tensor = np.zeros(list(pos.shape) + [4])\n",
    "neg_tensor = np.zeros(list(neg.shape) + [4])\n",
    "\n",
    "test_pos_tensor = np.zeros(list(test_pos.shape) + [4])\n",
    "test_neg_tensor = np.zeros(list(test_neg.shape) + [4])\n",
    "\n",
    "base_dict = {'A': 0, 'C': 1, 'G': 2, 'T': 3}\n",
    "\n",
    "#naive one-hot encoding\n",
    "for row in range(num_sample):\n",
    "    for col in range(seq_length):\n",
    "        pos_tensor[row,col,base_dict[pos[row,col]]] = 1\n",
    "        neg_tensor[row,col,base_dict[neg[row,col]]] = 1\n",
    "        if(row<test_size):\n",
    "            test_pos_tensor[row,col,base_dict[test_pos[row,col]]] = 1\n",
    "            test_neg_tensor[row,col,base_dict[test_neg[row,col]]] = 1        \n",
    "\n",
    "print('Positive sample matrix shape: {}'.format(pos.shape))\n",
    "# this should be a 3D tensor with shape: (samples, steps, input_dim)\n",
    "print('Positive sample tensor shape: {}'.format(pos_tensor.shape))\n",
    "\n",
    "X = np.vstack((pos_tensor, neg_tensor))\n",
    "y = np.concatenate((np.ones(num_sample), np.zeros(num_sample)))\n",
    "#y = np.concatenate((np.zeros(num_sample), np.ones(num_sample)))\n",
    "\n",
    "y_test = np.concatenate((np.ones(test_size), np.zeros(test_size)))\n",
    "#y_test = np.concatenate((np.zeros(test_size), np.ones(test_size)))\n",
    "\n",
    "print('Training set shape: {}'.format(X.shape))\n",
    "print('Training set label shape: {}'.format(y.shape))\n",
    "print('\\nOne-hot encoding looks like:\\n {}'.format(X[0,0:10,:]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(18000, 41, 4)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(18000,)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define Neural Network using Keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 51
    },
    "colab_type": "code",
    "id": "bD1LpHUKNtVh",
    "outputId": "ec377700-10d7-44a3-c80b-45c2a7133a11"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "#here comes the deep learning part\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv1D, Dense, Flatten, Dropout\n",
    "from keras.activations import relu\n",
    "from keras.layers.pooling import MaxPooling1D\n",
    "from keras.optimizers import SGD, Adam\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Flatten())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.add(Dense(41, activation='relu',input_shape=(41,4)))\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Dense(20, activation='relu'))\n",
    "model.add(Dropout(0.25))\n",
    "model.add(Dense(1, activation='sigmoid'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model.compile(optimizer=Adam(lr=0.001),loss='binary_crossentropy',metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 14400 samples, validate on 3600 samples\n",
      "Epoch 1/20\n",
      "14400/14400 [==============================] - 1s 63us/step - loss: 0.1786 - acc: 0.9241 - val_loss: 0.0442 - val_acc: 0.9839\n",
      "Epoch 2/20\n",
      "14400/14400 [==============================] - 1s 44us/step - loss: 0.0270 - acc: 0.9930 - val_loss: 0.0533 - val_acc: 0.9839\n",
      "Epoch 3/20\n",
      "14400/14400 [==============================] - 1s 37us/step - loss: 0.0162 - acc: 0.9965 - val_loss: 0.0456 - val_acc: 0.9883\n",
      "Epoch 4/20\n",
      "14400/14400 [==============================] - 1s 38us/step - loss: 0.0117 - acc: 0.9976 - val_loss: 0.0308 - val_acc: 0.9931\n",
      "Epoch 5/20\n",
      "14400/14400 [==============================] - 1s 37us/step - loss: 0.0104 - acc: 0.9978 - val_loss: 0.0535 - val_acc: 0.9900\n",
      "Epoch 6/20\n",
      "14400/14400 [==============================] - 1s 37us/step - loss: 0.0077 - acc: 0.9984 - val_loss: 0.0704 - val_acc: 0.9847\n",
      "Epoch 7/20\n",
      "14400/14400 [==============================] - 1s 37us/step - loss: 0.0046 - acc: 0.9993 - val_loss: 0.0709 - val_acc: 0.9872\n",
      "Epoch 8/20\n",
      "14400/14400 [==============================] - 1s 37us/step - loss: 0.0055 - acc: 0.9990 - val_loss: 0.0414 - val_acc: 0.9925\n",
      "Epoch 9/20\n",
      "14400/14400 [==============================] - 1s 39us/step - loss: 0.0052 - acc: 0.9989 - val_loss: 0.0632 - val_acc: 0.9897\n",
      "Epoch 10/20\n",
      "14400/14400 [==============================] - 1s 38us/step - loss: 0.0055 - acc: 0.9986 - val_loss: 0.0499 - val_acc: 0.9911\n",
      "Epoch 11/20\n",
      "14400/14400 [==============================] - 1s 38us/step - loss: 0.0041 - acc: 0.9991 - val_loss: 0.0545 - val_acc: 0.9914\n",
      "Epoch 12/20\n",
      "14400/14400 [==============================] - 1s 37us/step - loss: 0.0031 - acc: 0.9993 - val_loss: 0.0741 - val_acc: 0.9894\n",
      "Epoch 13/20\n",
      "14400/14400 [==============================] - 1s 37us/step - loss: 0.0036 - acc: 0.9991 - val_loss: 0.0777 - val_acc: 0.9883\n",
      "Epoch 14/20\n",
      "14400/14400 [==============================] - 1s 37us/step - loss: 0.0034 - acc: 0.9991 - val_loss: 0.0544 - val_acc: 0.9933\n",
      "Epoch 15/20\n",
      "14400/14400 [==============================] - 1s 39us/step - loss: 0.0027 - acc: 0.9994 - val_loss: 0.0679 - val_acc: 0.9914\n",
      "Epoch 16/20\n",
      "14400/14400 [==============================] - 1s 39us/step - loss: 0.0013 - acc: 0.9997 - val_loss: 0.0524 - val_acc: 0.9939\n",
      "Epoch 17/20\n",
      "14400/14400 [==============================] - 1s 47us/step - loss: 0.0023 - acc: 0.9993 - val_loss: 0.0692 - val_acc: 0.9911\n",
      "Epoch 18/20\n",
      "14400/14400 [==============================] - 1s 42us/step - loss: 0.0021 - acc: 0.9994 - val_loss: 0.0531 - val_acc: 0.9933\n",
      "Epoch 19/20\n",
      "14400/14400 [==============================] - 1s 44us/step - loss: 0.0018 - acc: 0.9995 - val_loss: 0.0725 - val_acc: 0.9914\n",
      "Epoch 20/20\n",
      "14400/14400 [==============================] - 1s 47us/step - loss: 0.0013 - acc: 0.9998 - val_loss: 0.0866 - val_acc: 0.9897\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X, y, validation_split=0.2, epochs=20)  # starts training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\nconvlayer = model.layers[0]\\nweights = convlayer.get_weights()[0].squeeze()\\nprint(\\'Convolution parameter shape: {}\\'.format(weights.shape))\\n\\nnum2seq = [\\'A\\',\\'C\\',\\'G\\',\\'T\\']\\n\\ntest_pred = \\'\\'.join([num2seq[np.argmax(weights[i,:])] for i in range(weights.shape[0])])\\n\\n\\n# characters stacked at the top of the RNAP PWM logo\\ntop_logo_chars = [\\'C\\',\\'T\\',\\'A\\',\\'C\\',\\'T\\',\\'A\\',\\'C\\',\\'A\\',\\'C\\',\\'G\\',\\n                  \\'C\\',\\'C\\',\\'C\\',\\'C\\',\\'A\\',\\'T\\',\\'A\\',\\'T\\',\\'T\\',\\'G\\',\\n                  \\'A\\',\\'G\\',\\'C\\',\\'G\\',\\'A\\',\\'T\\',\\'C\\',\\'C\\',\\'C\\',\\'G\\',\\n                  \\'G\\',\\'G\\',\\'T\\',\\'T\\',\\'C\\',\\'C\\',\\'C\\',\\'C\\',\\'C\\',\\'C\\',\\n                  \\'G\\']\\n\\ntest_pred_array = np.array(list(test_pred))\\ntest_pred_array\\n\\ntotal = 0\\nfor base_index in range(len(np.array(top_logo_chars))):\\n    if(top_logo_chars[base_index]==test_pred_array[base_index]):\\n        total+=1\\nprint(\"PWM top base prediction: \",total,\"out of 41\")  \\n'"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# I want to see if I can train the weights of the first conv layer to recover the \n",
    "# RNAP position weight matrix. I decided on using 1 filter with kernel size equal to \n",
    "# the sequences length (41). I will have to play around it with it more, right now the predictions\n",
    "# for the PWM don't look better than random ,on average, although sometimes I can get more than 60%\n",
    "# accuracy, though that might just be chance.\n",
    "\n",
    "\"\"\"\n",
    "convlayer = model.layers[0]\n",
    "weights = convlayer.get_weights()[0].squeeze()\n",
    "print('Convolution parameter shape: {}'.format(weights.shape))\n",
    "\n",
    "num2seq = ['A','C','G','T']\n",
    "\n",
    "test_pred = ''.join([num2seq[np.argmax(weights[i,:])] for i in range(weights.shape[0])])\n",
    "\n",
    "\n",
    "# characters stacked at the top of the RNAP PWM logo\n",
    "top_logo_chars = ['C','T','A','C','T','A','C','A','C','G',\n",
    "                  'C','C','C','C','A','T','A','T','T','G',\n",
    "                  'A','G','C','G','A','T','C','C','C','G',\n",
    "                  'G','G','T','T','C','C','C','C','C','C',\n",
    "                  'G']\n",
    "\n",
    "test_pred_array = np.array(list(test_pred))\n",
    "test_pred_array\n",
    "\n",
    "total = 0\n",
    "for base_index in range(len(np.array(top_logo_chars))):\n",
    "    if(top_logo_chars[base_index]==test_pred_array[base_index]):\n",
    "        total+=1\n",
    "print(\"PWM top base prediction: \",total,\"out of 41\")  \n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['val_loss', 'val_acc', 'loss', 'acc'])"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "history.history.keys()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Plot Loss functions and model Accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 376
    },
    "colab_type": "code",
    "id": "-VpzZqooOAPx",
    "outputId": "30d34bfd-b21e-47c7-9c64-d1cf407b23c6"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.figure()\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 376
    },
    "colab_type": "code",
    "id": "VeIdsrBEO49U",
    "outputId": "04003ed6-30d1-494a-da95-3672e19fb46d"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nHx+4vlHaEtNnAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure()\n",
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Make predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y_pred_pos = model.predict(test_pos_tensor)\n",
    "y_pred_neg = model.predict(test_neg_tensor)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAD8CAYAAAB+UHOxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIhIoBYCISKD+P8m+yuqFVxSNAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(y_pred_pos,bins=100,label='positive_test')\n",
    "plt.hist(y_pred_neg,bins=100,label='negative_test')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "wrong_neg_predictions = 0\n",
    "for neg_index in range(len(y_test[1000:2000].reshape(1000,1))):\n",
    "    if(abs(y_pred_neg[neg_index]-y_test[neg_index+1000])>0.1):\n",
    "        #print(neg_index,y_pred_neg[neg_index])\n",
    "        wrong_neg_predictions+=1\n",
    "\n",
    "wrong_pos_predictions = 0\n",
    "for pos_index in range(len(y_test[0:1000].reshape(1000,1))):\n",
    "    if(abs(y_pred_pos[pos_index]-y_test[pos_index])>0.1):\n",
    "        #print(pos_index,y_pred_pos[pos_index])\n",
    "        wrong_pos_predictions+=1\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "wrong predictions on negative test: 8 out of  1000  predictions\n",
      "wrong predictions on positive test: 0  out of  1000  predictions\n"
     ]
    }
   ],
   "source": [
    "print(\"wrong predictions on negative test:\",wrong_neg_predictions, \"out of \",len(y_pred_pos),\" predictions\")\n",
    "print(\"wrong predictions on positive test:\",wrong_pos_predictions,\" out of \",len(y_pred_neg),\" predictions\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.999997  ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999989 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999883 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999862 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [0.9999999 ],\n",
       "       [0.9999999 ],\n",
       "       [0.9999999 ],\n",
       "       [0.9999999 ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999989 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999925 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999905],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999993 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.999995  ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999982 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99998236],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999399 ],\n",
       "       [1.        ],\n",
       "       [0.9999995 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9998248 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999827 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999951 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999535],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999964],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999788 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99998367],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999995 ],\n",
       "       [0.9999989 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999945 ],\n",
       "       [0.9999958 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999917],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999905],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [0.9999974 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [0.9999993 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999154],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.999936  ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999917],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999976 ],\n",
       "       [1.        ],\n",
       "       [0.9999987 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999523],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99967766],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99998164],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.99999976],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [0.99999917],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999999 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [0.9999689 ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ],\n",
       "       [1.        ]], dtype=float32)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_pos"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "motif_learn_convnet.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}