perceptron_synt.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "import sklearn.model_selection\n",
    "import numpy as np\n",
    "\n",
    "\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create synthetic dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Instead of using the actual fingerpint data, this is an example with \"fake\" fingerprints.\n",
    "I create these as follows: \n",
    "* the fingerprints are 1000 bits long.\n",
    "* For active molecules, the first 10% of the bits is mostly \"on\", and the last third is mostly \"off\".\n",
    "* for inactive molecules, the opposite: the first 10%% is mostly off and the last mostly on.\n",
    "* the middle 80% is just noise, on/off are equally probable for both sets.\n",
    "\n",
    "100 ligands in each set (active and inactive), 200 in total"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training/validation dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "fp_len = 1000\n",
    "meaningul_frac = 0.10\n",
    "\n",
    "## what is the probability of having an on bit in the meaninful region for actives?\n",
    "pact=0.65\n",
    "pinact=1-pact\n",
    "\n",
    "## how many ligands in each set?\n",
    "n_act=100\n",
    "n_inact=100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "act=[]\n",
    "inact=[]\n",
    "\n",
    "## generate the active\n",
    "for i in range(n_act):\n",
    "    odds = np.random.binomial(1,pact,size=int(fp_len*meaningul_frac))\n",
    "    evens = np.random.binomial(1,1.-pact,size=int(fp_len*meaningul_frac))\n",
    "    noise = np.random.binomial(1,0.5,size=int(fp_len*(1-2*meaningul_frac)))\n",
    "    out = np.array([1])\n",
    "    act.append(np.concatenate((odds,noise,evens,out)))\n",
    "\n",
    "## generate the inactive\n",
    "for i in range(n_inact):\n",
    "    odds = np.random.binomial(1,pinact,size=int(fp_len*meaningul_frac))\n",
    "    evens = np.random.binomial(1,1.-pinact,size=int(fp_len*meaningul_frac))\n",
    "    noise = np.random.binomial(1,0.5,size=int(fp_len*(1-2*meaningul_frac)))\n",
    "    out = np.array([1])\n",
    "    out = np.array([0])\n",
    "    inact.append(np.concatenate((odds,noise,evens,out)))\n",
    "\n",
    "## put them together; x_all are the input for the network, y_all are the output\n",
    "## which transform to categorical output\n",
    "\n",
    "data = np.vstack(act+inact)\n",
    "x_all = data[:,0:-1]\n",
    "y_all = keras.utils.np_utils.to_categorical(data[:,-1])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.lines.Line2D at 0x7f432549e358>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_all)\n",
    "plt.ylabel('ligands')\n",
    "plt.xlabel('bits of fingerprint')\n",
    "plt.axvline(x=int(fp_len*meaningul_frac),color='red')\n",
    "plt.axvline(x=int(fp_len*(1-meaningul_frac)),color='red')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "the training/validation set is split randomly in 30% for validation, 70% for training\n",
    "the validation set is used as follows:\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train, x_valid, y_train, y_valid = sklearn.model_selection.train_test_split(\n",
    "    x_all, \n",
    "    y_all, \n",
    "    test_size=0.3, \n",
    "    random_state=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### test set\n",
    "I create a new, independend set using the same definitions as the training to test the predictions of our model.\n",
    "once the neural network has learned the training set, we use the test set to see how well it works"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "## geenrate test set:\n",
    "act=[]\n",
    "inact=[]\n",
    "\n",
    "## generate the active\n",
    "for i in range(n_act):\n",
    "    odds = np.random.binomial(1,pact,size=int(fp_len*meaningul_frac))\n",
    "    evens = np.random.binomial(1,1.-pact,size=int(fp_len*meaningul_frac))\n",
    "    noise = np.random.binomial(1,0.5,size=int(fp_len*(1-2*meaningul_frac)))\n",
    "    out = np.array([1])\n",
    "    act.append(np.concatenate((odds,noise,evens,out)))\n",
    "\n",
    "## generate the inactive\n",
    "for i in range(n_inact):\n",
    "    odds = np.random.binomial(1,pinact,size=int(fp_len*meaningul_frac))\n",
    "    evens = np.random.binomial(1,1.-pinact,size=int(fp_len*meaningul_frac))\n",
    "    noise = np.random.binomial(1,0.5,size=int(fp_len*(1-2*meaningul_frac)))\n",
    "    out = np.array([0])\n",
    "    inact.append(np.concatenate((odds,noise,evens,out)))\n",
    "\n",
    "## put them together; x_all are the input for the network, y_all are the output\n",
    "## which transform to categorical output\n",
    "\n",
    "data_test = np.vstack(act+inact)\n",
    "x_test = data_test[:,0:-1]\n",
    "y_test_wanted = data_test[:,-1]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define the neural network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "batch_normalization_13 (Batc (None, 1000)              4000      \n",
      "_________________________________________________________________\n",
      "dense_25 (Dense)             (None, 32)                32032     \n",
      "_________________________________________________________________\n",
      "dense_26 (Dense)             (None, 16)                528       \n",
      "_________________________________________________________________\n",
      "dense_27 (Dense)             (None, 16)                272       \n",
      "_________________________________________________________________\n",
      "dense_28 (Dense)             (None, 2)                 34        \n",
      "=================================================================\n",
      "Total params: 36,866\n",
      "Trainable params: 34,866\n",
      "Non-trainable params: 2,000\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model = keras.models.Sequential([\n",
    "    keras.layers.normalization.BatchNormalization(input_shape=tuple([x_train.shape[1]])),\n",
    "    keras.layers.core.Dense(32, activation='relu'),\n",
    "    keras.layers.core.Dense(16, activation='relu'),\n",
    "    keras.layers.core.Dense(16, activation='relu'),\n",
    "    keras.layers.core.Dense(2,   activation='sigmoid')\n",
    "])\n",
    "\n",
    "\n",
    "#model.compile(\n",
    "#    loss=\"categorical_crossentropy\", \n",
    "#    optimizer=\"adadelta\",\n",
    "#    metrics=[\"accuracy\"])\n",
    "\n",
    "model.compile(keras.optimizers.Adam(lr=0.001), \n",
    "              loss='categorical_crossentropy', \n",
    "              metrics=[\"accuracy\"])\n",
    "\n",
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "callback_early_stopping = keras.callbacks.EarlyStopping(\n",
    "    monitor='val_loss', \n",
    "    patience=10, \n",
    "    verbose=0, \n",
    "    mode='auto')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## fit the neural network using the test set and the validation set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 140 samples, validate on 60 samples\n",
      "Epoch 1/200\n",
      " - 0s - loss: 3.1710e-04 - acc: 1.0000 - val_loss: 0.0236 - val_acc: 1.0000\n",
      "Epoch 2/200\n",
      " - 0s - loss: 2.8986e-04 - acc: 1.0000 - val_loss: 0.0234 - val_acc: 1.0000\n",
      "Epoch 3/200\n",
      " - 0s - loss: 4.1825e-04 - acc: 1.0000 - val_loss: 0.0232 - val_acc: 1.0000\n",
      "Epoch 4/200\n",
      " - 0s - loss: 2.8706e-04 - acc: 1.0000 - val_loss: 0.0231 - val_acc: 1.0000\n",
      "Epoch 5/200\n",
      " - 0s - loss: 3.1075e-04 - acc: 1.0000 - val_loss: 0.0230 - val_acc: 1.0000\n",
      "Epoch 6/200\n",
      " - 0s - loss: 3.3211e-04 - acc: 1.0000 - val_loss: 0.0229 - val_acc: 1.0000\n",
      "Epoch 7/200\n",
      " - 0s - loss: 2.7067e-04 - acc: 1.0000 - val_loss: 0.0228 - val_acc: 1.0000\n",
      "Epoch 8/200\n",
      " - 0s - loss: 3.0369e-04 - acc: 1.0000 - val_loss: 0.0227 - val_acc: 1.0000\n",
      "Epoch 9/200\n",
      " - 0s - loss: 7.8615e-04 - acc: 1.0000 - val_loss: 0.0224 - val_acc: 1.0000\n",
      "Epoch 10/200\n",
      " - 0s - loss: 2.5842e-04 - acc: 1.0000 - val_loss: 0.0223 - val_acc: 1.0000\n",
      "Epoch 11/200\n",
      " - 0s - loss: 3.1776e-04 - acc: 1.0000 - val_loss: 0.0220 - val_acc: 1.0000\n",
      "Epoch 12/200\n",
      " - 0s - loss: 2.4895e-04 - acc: 1.0000 - val_loss: 0.0219 - val_acc: 1.0000\n",
      "Epoch 13/200\n",
      " - 0s - loss: 2.4477e-04 - acc: 1.0000 - val_loss: 0.0218 - val_acc: 1.0000\n",
      "Epoch 14/200\n",
      " - 0s - loss: 2.3550e-04 - acc: 1.0000 - val_loss: 0.0217 - val_acc: 1.0000\n",
      "Epoch 15/200\n",
      " - 0s - loss: 3.5971e-04 - acc: 1.0000 - val_loss: 0.0215 - val_acc: 1.0000\n",
      "Epoch 16/200\n",
      " - 0s - loss: 2.2364e-04 - acc: 1.0000 - val_loss: 0.0214 - val_acc: 1.0000\n",
      "Epoch 17/200\n",
      " - 0s - loss: 2.3213e-04 - acc: 1.0000 - val_loss: 0.0212 - val_acc: 1.0000\n",
      "Epoch 18/200\n",
      " - 0s - loss: 2.3540e-04 - acc: 1.0000 - val_loss: 0.0212 - val_acc: 1.0000\n",
      "Epoch 19/200\n",
      " - 0s - loss: 2.6932e-04 - acc: 1.0000 - val_loss: 0.0210 - val_acc: 1.0000\n",
      "Epoch 20/200\n",
      " - 0s - loss: 2.2474e-04 - acc: 1.0000 - val_loss: 0.0209 - val_acc: 1.0000\n",
      "Epoch 21/200\n",
      " - 0s - loss: 2.0439e-04 - acc: 1.0000 - val_loss: 0.0207 - val_acc: 1.0000\n",
      "Epoch 22/200\n",
      " - 0s - loss: 2.3457e-04 - acc: 1.0000 - val_loss: 0.0207 - val_acc: 1.0000\n",
      "Epoch 23/200\n",
      " - 0s - loss: 1.9902e-04 - acc: 1.0000 - val_loss: 0.0206 - val_acc: 1.0000\n",
      "Epoch 24/200\n",
      " - 0s - loss: 2.1203e-04 - acc: 1.0000 - val_loss: 0.0204 - val_acc: 1.0000\n",
      "Epoch 25/200\n",
      " - 0s - loss: 2.1475e-04 - acc: 1.0000 - val_loss: 0.0203 - val_acc: 1.0000\n",
      "Epoch 26/200\n",
      " - 0s - loss: 1.9577e-04 - acc: 1.0000 - val_loss: 0.0203 - val_acc: 1.0000\n",
      "Epoch 27/200\n",
      " - 0s - loss: 1.9054e-04 - acc: 1.0000 - val_loss: 0.0202 - val_acc: 1.0000\n",
      "Epoch 28/200\n",
      " - 0s - loss: 1.8744e-04 - acc: 1.0000 - val_loss: 0.0201 - val_acc: 1.0000\n",
      "Epoch 29/200\n",
      " - 0s - loss: 1.8939e-04 - acc: 1.0000 - val_loss: 0.0199 - val_acc: 1.0000\n",
      "Epoch 30/200\n",
      " - 0s - loss: 1.9102e-04 - acc: 1.0000 - val_loss: 0.0198 - val_acc: 1.0000\n",
      "Epoch 31/200\n",
      " - 0s - loss: 2.3015e-04 - acc: 1.0000 - val_loss: 0.0196 - val_acc: 1.0000\n",
      "Epoch 32/200\n",
      " - 0s - loss: 2.0778e-04 - acc: 1.0000 - val_loss: 0.0196 - val_acc: 1.0000\n",
      "Epoch 33/200\n",
      " - 0s - loss: 1.9309e-04 - acc: 1.0000 - val_loss: 0.0194 - val_acc: 1.0000\n",
      "Epoch 34/200\n",
      " - 0s - loss: 1.9034e-04 - acc: 1.0000 - val_loss: 0.0193 - val_acc: 1.0000\n",
      "Epoch 35/200\n",
      " - 0s - loss: 3.0052e-04 - acc: 1.0000 - val_loss: 0.0193 - val_acc: 1.0000\n",
      "Epoch 36/200\n",
      " - 0s - loss: 1.6424e-04 - acc: 1.0000 - val_loss: 0.0191 - val_acc: 1.0000\n",
      "Epoch 37/200\n",
      " - 0s - loss: 1.6684e-04 - acc: 1.0000 - val_loss: 0.0191 - val_acc: 1.0000\n",
      "Epoch 38/200\n",
      " - 0s - loss: 1.7153e-04 - acc: 1.0000 - val_loss: 0.0189 - val_acc: 1.0000\n",
      "Epoch 39/200\n",
      " - 0s - loss: 1.5697e-04 - acc: 1.0000 - val_loss: 0.0188 - val_acc: 1.0000\n",
      "Epoch 40/200\n",
      " - 0s - loss: 1.5246e-04 - acc: 1.0000 - val_loss: 0.0187 - val_acc: 1.0000\n",
      "Epoch 41/200\n",
      " - 0s - loss: 1.8834e-04 - acc: 1.0000 - val_loss: 0.0186 - val_acc: 1.0000\n",
      "Epoch 42/200\n",
      " - 0s - loss: 1.6863e-04 - acc: 1.0000 - val_loss: 0.0185 - val_acc: 1.0000\n",
      "Epoch 43/200\n",
      " - 0s - loss: 1.6701e-04 - acc: 1.0000 - val_loss: 0.0183 - val_acc: 1.0000\n",
      "Epoch 44/200\n",
      " - 0s - loss: 2.4709e-04 - acc: 1.0000 - val_loss: 0.0184 - val_acc: 1.0000\n",
      "Epoch 45/200\n",
      " - 0s - loss: 1.4412e-04 - acc: 1.0000 - val_loss: 0.0183 - val_acc: 1.0000\n",
      "Epoch 46/200\n",
      " - 0s - loss: 1.4029e-04 - acc: 1.0000 - val_loss: 0.0182 - val_acc: 1.0000\n",
      "Epoch 47/200\n",
      " - 0s - loss: 1.4920e-04 - acc: 1.0000 - val_loss: 0.0181 - val_acc: 1.0000\n",
      "Epoch 48/200\n",
      " - 0s - loss: 1.3878e-04 - acc: 1.0000 - val_loss: 0.0181 - val_acc: 1.0000\n",
      "Epoch 49/200\n",
      " - 0s - loss: 1.3273e-04 - acc: 1.0000 - val_loss: 0.0180 - val_acc: 1.0000\n",
      "Epoch 50/200\n",
      " - 0s - loss: 1.4199e-04 - acc: 1.0000 - val_loss: 0.0178 - val_acc: 1.0000\n",
      "Epoch 51/200\n",
      " - 0s - loss: 2.2985e-04 - acc: 1.0000 - val_loss: 0.0179 - val_acc: 1.0000\n",
      "Epoch 52/200\n",
      " - 0s - loss: 1.2691e-04 - acc: 1.0000 - val_loss: 0.0178 - val_acc: 1.0000\n",
      "Epoch 53/200\n",
      " - 0s - loss: 1.2858e-04 - acc: 1.0000 - val_loss: 0.0177 - val_acc: 1.0000\n",
      "Epoch 54/200\n",
      " - 0s - loss: 1.2422e-04 - acc: 1.0000 - val_loss: 0.0176 - val_acc: 1.0000\n",
      "Epoch 55/200\n",
      " - 0s - loss: 1.5396e-04 - acc: 1.0000 - val_loss: 0.0174 - val_acc: 1.0000\n",
      "Epoch 56/200\n",
      " - 0s - loss: 1.2922e-04 - acc: 1.0000 - val_loss: 0.0173 - val_acc: 1.0000\n",
      "Epoch 57/200\n",
      " - 0s - loss: 1.2650e-04 - acc: 1.0000 - val_loss: 0.0173 - val_acc: 1.0000\n",
      "Epoch 58/200\n",
      " - 0s - loss: 1.2431e-04 - acc: 1.0000 - val_loss: 0.0172 - val_acc: 1.0000\n",
      "Epoch 59/200\n",
      " - 0s - loss: 1.2253e-04 - acc: 1.0000 - val_loss: 0.0171 - val_acc: 1.0000\n",
      "Epoch 60/200\n",
      " - 0s - loss: 1.1529e-04 - acc: 1.0000 - val_loss: 0.0170 - val_acc: 1.0000\n",
      "Epoch 61/200\n",
      " - 0s - loss: 1.1855e-04 - acc: 1.0000 - val_loss: 0.0169 - val_acc: 1.0000\n",
      "Epoch 62/200\n",
      " - 0s - loss: 1.2503e-04 - acc: 1.0000 - val_loss: 0.0168 - val_acc: 1.0000\n",
      "Epoch 63/200\n",
      " - 0s - loss: 1.1744e-04 - acc: 1.0000 - val_loss: 0.0168 - val_acc: 1.0000\n",
      "Epoch 64/200\n",
      " - 0s - loss: 1.1092e-04 - acc: 1.0000 - val_loss: 0.0167 - val_acc: 1.0000\n",
      "Epoch 65/200\n",
      " - 0s - loss: 1.4079e-04 - acc: 1.0000 - val_loss: 0.0166 - val_acc: 1.0000\n",
      "Epoch 66/200\n",
      " - 0s - loss: 1.1133e-04 - acc: 1.0000 - val_loss: 0.0165 - val_acc: 1.0000\n",
      "Epoch 67/200\n",
      " - 0s - loss: 1.1099e-04 - acc: 1.0000 - val_loss: 0.0164 - val_acc: 1.0000\n",
      "Epoch 68/200\n",
      " - 0s - loss: 1.0186e-04 - acc: 1.0000 - val_loss: 0.0164 - val_acc: 1.0000\n",
      "Epoch 69/200\n",
      " - 0s - loss: 1.4035e-04 - acc: 1.0000 - val_loss: 0.0163 - val_acc: 1.0000\n",
      "Epoch 70/200\n",
      " - 0s - loss: 1.1938e-04 - acc: 1.0000 - val_loss: 0.0163 - val_acc: 1.0000\n",
      "Epoch 71/200\n",
      " - 0s - loss: 1.0498e-04 - acc: 1.0000 - val_loss: 0.0163 - val_acc: 1.0000\n",
      "Epoch 72/200\n",
      " - 0s - loss: 1.0498e-04 - acc: 1.0000 - val_loss: 0.0162 - val_acc: 1.0000\n",
      "Epoch 73/200\n",
      " - 0s - loss: 9.8666e-05 - acc: 1.0000 - val_loss: 0.0162 - val_acc: 1.0000\n",
      "Epoch 74/200\n",
      " - 0s - loss: 1.0053e-04 - acc: 1.0000 - val_loss: 0.0162 - val_acc: 1.0000\n",
      "Epoch 75/200\n",
      " - 0s - loss: 1.1061e-04 - acc: 1.0000 - val_loss: 0.0161 - val_acc: 1.0000\n",
      "Epoch 76/200\n",
      " - 0s - loss: 1.0591e-04 - acc: 1.0000 - val_loss: 0.0161 - val_acc: 1.0000\n",
      "Epoch 77/200\n",
      " - 0s - loss: 9.8451e-05 - acc: 1.0000 - val_loss: 0.0161 - val_acc: 1.0000\n",
      "Epoch 78/200\n",
      " - 0s - loss: 1.2669e-04 - acc: 1.0000 - val_loss: 0.0159 - val_acc: 1.0000\n",
      "Epoch 79/200\n",
      " - 0s - loss: 9.7473e-05 - acc: 1.0000 - val_loss: 0.0159 - val_acc: 1.0000\n",
      "Epoch 80/200\n",
      " - 0s - loss: 1.1473e-04 - acc: 1.0000 - val_loss: 0.0158 - val_acc: 1.0000\n",
      "Epoch 81/200\n",
      " - 0s - loss: 8.9512e-05 - acc: 1.0000 - val_loss: 0.0157 - val_acc: 1.0000\n",
      "Epoch 82/200\n",
      " - 0s - loss: 1.2437e-04 - acc: 1.0000 - val_loss: 0.0156 - val_acc: 1.0000\n",
      "Epoch 83/200\n",
      " - 0s - loss: 1.0362e-04 - acc: 1.0000 - val_loss: 0.0155 - val_acc: 1.0000\n",
      "Epoch 84/200\n",
      " - 0s - loss: 1.2371e-04 - acc: 1.0000 - val_loss: 0.0154 - val_acc: 1.0000\n",
      "Epoch 85/200\n",
      " - 0s - loss: 8.5998e-05 - acc: 1.0000 - val_loss: 0.0154 - val_acc: 1.0000\n",
      "Epoch 86/200\n",
      " - 0s - loss: 9.4891e-05 - acc: 1.0000 - val_loss: 0.0154 - val_acc: 1.0000\n",
      "Epoch 87/200\n",
      " - 0s - loss: 1.0408e-04 - acc: 1.0000 - val_loss: 0.0152 - val_acc: 1.0000\n",
      "Epoch 88/200\n",
      " - 0s - loss: 1.0205e-04 - acc: 1.0000 - val_loss: 0.0152 - val_acc: 1.0000\n",
      "Epoch 89/200\n",
      " - 0s - loss: 1.4745e-04 - acc: 1.0000 - val_loss: 0.0153 - val_acc: 1.0000\n",
      "Epoch 90/200\n",
      " - 0s - loss: 1.2477e-04 - acc: 1.0000 - val_loss: 0.0152 - val_acc: 1.0000\n",
      "Epoch 91/200\n",
      " - 0s - loss: 8.7136e-05 - acc: 1.0000 - val_loss: 0.0151 - val_acc: 1.0000\n",
      "Epoch 92/200\n",
      " - 0s - loss: 9.3960e-05 - acc: 1.0000 - val_loss: 0.0151 - val_acc: 1.0000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 93/200\n",
      " - 0s - loss: 7.8719e-05 - acc: 1.0000 - val_loss: 0.0150 - val_acc: 1.0000\n",
      "Epoch 94/200\n",
      " - 0s - loss: 8.3706e-05 - acc: 1.0000 - val_loss: 0.0150 - val_acc: 1.0000\n",
      "Epoch 95/200\n",
      " - 0s - loss: 8.3737e-05 - acc: 1.0000 - val_loss: 0.0149 - val_acc: 1.0000\n",
      "Epoch 96/200\n",
      " - 0s - loss: 7.9502e-05 - acc: 1.0000 - val_loss: 0.0149 - val_acc: 1.0000\n",
      "Epoch 97/200\n",
      " - 0s - loss: 2.2958e-04 - acc: 1.0000 - val_loss: 0.0148 - val_acc: 1.0000\n",
      "Epoch 98/200\n",
      " - 0s - loss: 1.0507e-04 - acc: 1.0000 - val_loss: 0.0148 - val_acc: 1.0000\n",
      "Epoch 99/200\n",
      " - 0s - loss: 8.9041e-05 - acc: 1.0000 - val_loss: 0.0147 - val_acc: 1.0000\n",
      "Epoch 100/200\n",
      " - 0s - loss: 8.9843e-05 - acc: 1.0000 - val_loss: 0.0146 - val_acc: 1.0000\n",
      "Epoch 101/200\n",
      " - 0s - loss: 8.5042e-05 - acc: 1.0000 - val_loss: 0.0146 - val_acc: 1.0000\n",
      "Epoch 102/200\n",
      " - 0s - loss: 8.3304e-05 - acc: 1.0000 - val_loss: 0.0146 - val_acc: 1.0000\n",
      "Epoch 103/200\n",
      " - 0s - loss: 1.2345e-04 - acc: 1.0000 - val_loss: 0.0145 - val_acc: 1.0000\n",
      "Epoch 104/200\n",
      " - 0s - loss: 1.2921e-04 - acc: 1.0000 - val_loss: 0.0144 - val_acc: 1.0000\n",
      "Epoch 105/200\n",
      " - 0s - loss: 7.5587e-05 - acc: 1.0000 - val_loss: 0.0144 - val_acc: 1.0000\n",
      "Epoch 106/200\n",
      " - 0s - loss: 6.8241e-05 - acc: 1.0000 - val_loss: 0.0143 - val_acc: 1.0000\n",
      "Epoch 107/200\n",
      " - 0s - loss: 7.4073e-05 - acc: 1.0000 - val_loss: 0.0142 - val_acc: 1.0000\n",
      "Epoch 108/200\n",
      " - 0s - loss: 8.7805e-05 - acc: 1.0000 - val_loss: 0.0141 - val_acc: 1.0000\n",
      "Epoch 109/200\n",
      " - 0s - loss: 8.2042e-05 - acc: 1.0000 - val_loss: 0.0142 - val_acc: 1.0000\n",
      "Epoch 110/200\n",
      " - 0s - loss: 6.4912e-05 - acc: 1.0000 - val_loss: 0.0141 - val_acc: 1.0000\n",
      "Epoch 111/200\n",
      " - 0s - loss: 8.5892e-05 - acc: 1.0000 - val_loss: 0.0140 - val_acc: 1.0000\n",
      "Epoch 112/200\n",
      " - 0s - loss: 6.7725e-05 - acc: 1.0000 - val_loss: 0.0140 - val_acc: 1.0000\n",
      "Epoch 113/200\n",
      " - 0s - loss: 7.0865e-05 - acc: 1.0000 - val_loss: 0.0140 - val_acc: 1.0000\n",
      "Epoch 114/200\n",
      " - 0s - loss: 6.8159e-05 - acc: 1.0000 - val_loss: 0.0140 - val_acc: 1.0000\n",
      "Epoch 115/200\n",
      " - 0s - loss: 6.2870e-05 - acc: 1.0000 - val_loss: 0.0140 - val_acc: 1.0000\n",
      "Epoch 116/200\n",
      " - 0s - loss: 6.7454e-05 - acc: 1.0000 - val_loss: 0.0140 - val_acc: 1.0000\n",
      "Epoch 117/200\n",
      " - 0s - loss: 6.2842e-05 - acc: 1.0000 - val_loss: 0.0139 - val_acc: 1.0000\n",
      "Epoch 118/200\n",
      " - 0s - loss: 6.1013e-05 - acc: 1.0000 - val_loss: 0.0138 - val_acc: 1.0000\n",
      "Epoch 119/200\n",
      " - 0s - loss: 5.9413e-05 - acc: 1.0000 - val_loss: 0.0138 - val_acc: 1.0000\n",
      "Epoch 120/200\n",
      " - 0s - loss: 5.9389e-05 - acc: 1.0000 - val_loss: 0.0138 - val_acc: 1.0000\n",
      "Epoch 121/200\n",
      " - 0s - loss: 8.7526e-05 - acc: 1.0000 - val_loss: 0.0137 - val_acc: 1.0000\n",
      "Epoch 122/200\n",
      " - 0s - loss: 5.7229e-05 - acc: 1.0000 - val_loss: 0.0137 - val_acc: 1.0000\n",
      "Epoch 123/200\n",
      " - 0s - loss: 5.8222e-05 - acc: 1.0000 - val_loss: 0.0136 - val_acc: 1.0000\n",
      "Epoch 124/200\n",
      " - 0s - loss: 8.4266e-05 - acc: 1.0000 - val_loss: 0.0136 - val_acc: 1.0000\n",
      "Epoch 125/200\n",
      " - 0s - loss: 5.6078e-05 - acc: 1.0000 - val_loss: 0.0136 - val_acc: 1.0000\n",
      "Epoch 126/200\n",
      " - 0s - loss: 8.6878e-05 - acc: 1.0000 - val_loss: 0.0134 - val_acc: 1.0000\n",
      "Epoch 127/200\n",
      " - 0s - loss: 6.6030e-05 - acc: 1.0000 - val_loss: 0.0134 - val_acc: 1.0000\n",
      "Epoch 128/200\n",
      " - 0s - loss: 5.9917e-05 - acc: 1.0000 - val_loss: 0.0134 - val_acc: 1.0000\n",
      "Epoch 129/200\n",
      " - 0s - loss: 5.4062e-05 - acc: 1.0000 - val_loss: 0.0134 - val_acc: 1.0000\n",
      "Epoch 130/200\n",
      " - 0s - loss: 5.4348e-05 - acc: 1.0000 - val_loss: 0.0134 - val_acc: 1.0000\n",
      "Epoch 131/200\n",
      " - 0s - loss: 5.2947e-05 - acc: 1.0000 - val_loss: 0.0133 - val_acc: 1.0000\n",
      "Epoch 132/200\n",
      " - 0s - loss: 6.2798e-05 - acc: 1.0000 - val_loss: 0.0133 - val_acc: 1.0000\n",
      "Epoch 133/200\n",
      " - 0s - loss: 7.9740e-05 - acc: 1.0000 - val_loss: 0.0134 - val_acc: 1.0000\n",
      "Epoch 134/200\n",
      " - 0s - loss: 5.9936e-05 - acc: 1.0000 - val_loss: 0.0133 - val_acc: 1.0000\n",
      "Epoch 135/200\n",
      " - 0s - loss: 5.7778e-05 - acc: 1.0000 - val_loss: 0.0132 - val_acc: 1.0000\n",
      "Epoch 136/200\n",
      " - 0s - loss: 5.3287e-05 - acc: 1.0000 - val_loss: 0.0132 - val_acc: 1.0000\n",
      "Epoch 137/200\n",
      " - 0s - loss: 5.1839e-05 - acc: 1.0000 - val_loss: 0.0131 - val_acc: 1.0000\n",
      "Epoch 138/200\n",
      " - 0s - loss: 5.1266e-05 - acc: 1.0000 - val_loss: 0.0131 - val_acc: 1.0000\n",
      "Epoch 139/200\n",
      " - 0s - loss: 7.2415e-05 - acc: 1.0000 - val_loss: 0.0131 - val_acc: 1.0000\n",
      "Epoch 140/200\n",
      " - 0s - loss: 4.8974e-05 - acc: 1.0000 - val_loss: 0.0131 - val_acc: 1.0000\n",
      "Epoch 141/200\n",
      " - 0s - loss: 6.0115e-05 - acc: 1.0000 - val_loss: 0.0130 - val_acc: 1.0000\n",
      "Epoch 142/200\n",
      " - 0s - loss: 5.4633e-05 - acc: 1.0000 - val_loss: 0.0129 - val_acc: 1.0000\n",
      "Epoch 143/200\n",
      " - 0s - loss: 6.7615e-05 - acc: 1.0000 - val_loss: 0.0129 - val_acc: 1.0000\n",
      "Epoch 144/200\n",
      " - 0s - loss: 7.1305e-05 - acc: 1.0000 - val_loss: 0.0128 - val_acc: 1.0000\n",
      "Epoch 145/200\n",
      " - 0s - loss: 9.8786e-05 - acc: 1.0000 - val_loss: 0.0127 - val_acc: 1.0000\n",
      "Epoch 146/200\n",
      " - 0s - loss: 6.2614e-05 - acc: 1.0000 - val_loss: 0.0126 - val_acc: 1.0000\n",
      "Epoch 147/200\n",
      " - 0s - loss: 4.7818e-05 - acc: 1.0000 - val_loss: 0.0126 - val_acc: 1.0000\n",
      "Epoch 148/200\n",
      " - 0s - loss: 5.2608e-05 - acc: 1.0000 - val_loss: 0.0126 - val_acc: 1.0000\n",
      "Epoch 149/200\n",
      " - 0s - loss: 4.9296e-05 - acc: 1.0000 - val_loss: 0.0125 - val_acc: 1.0000\n",
      "Epoch 150/200\n",
      " - 0s - loss: 5.2791e-05 - acc: 1.0000 - val_loss: 0.0125 - val_acc: 1.0000\n",
      "Epoch 151/200\n",
      " - 0s - loss: 4.6672e-05 - acc: 1.0000 - val_loss: 0.0125 - val_acc: 1.0000\n",
      "Epoch 152/200\n",
      " - 0s - loss: 4.8344e-05 - acc: 1.0000 - val_loss: 0.0124 - val_acc: 1.0000\n",
      "Epoch 153/200\n",
      " - 0s - loss: 6.1394e-05 - acc: 1.0000 - val_loss: 0.0125 - val_acc: 1.0000\n",
      "Epoch 154/200\n",
      " - 0s - loss: 5.1500e-05 - acc: 1.0000 - val_loss: 0.0124 - val_acc: 1.0000\n",
      "Epoch 155/200\n",
      " - 0s - loss: 4.6727e-05 - acc: 1.0000 - val_loss: 0.0123 - val_acc: 1.0000\n",
      "Epoch 156/200\n",
      " - 0s - loss: 8.0575e-05 - acc: 1.0000 - val_loss: 0.0121 - val_acc: 1.0000\n",
      "Epoch 157/200\n",
      " - 0s - loss: 5.2495e-05 - acc: 1.0000 - val_loss: 0.0122 - val_acc: 1.0000\n",
      "Epoch 158/200\n",
      " - 0s - loss: 4.2698e-05 - acc: 1.0000 - val_loss: 0.0122 - val_acc: 1.0000\n",
      "Epoch 159/200\n",
      " - 0s - loss: 4.2785e-05 - acc: 1.0000 - val_loss: 0.0121 - val_acc: 1.0000\n",
      "Epoch 160/200\n",
      " - 0s - loss: 6.3418e-05 - acc: 1.0000 - val_loss: 0.0122 - val_acc: 1.0000\n",
      "Epoch 161/200\n",
      " - 0s - loss: 4.4265e-05 - acc: 1.0000 - val_loss: 0.0121 - val_acc: 1.0000\n",
      "Epoch 162/200\n",
      " - 0s - loss: 4.0286e-05 - acc: 1.0000 - val_loss: 0.0121 - val_acc: 1.0000\n",
      "Epoch 163/200\n",
      " - 0s - loss: 4.8397e-05 - acc: 1.0000 - val_loss: 0.0120 - val_acc: 1.0000\n",
      "Epoch 164/200\n",
      " - 0s - loss: 4.0662e-05 - acc: 1.0000 - val_loss: 0.0120 - val_acc: 1.0000\n",
      "Epoch 165/200\n",
      " - 0s - loss: 4.7095e-05 - acc: 1.0000 - val_loss: 0.0120 - val_acc: 1.0000\n",
      "Epoch 166/200\n",
      " - 0s - loss: 6.3094e-05 - acc: 1.0000 - val_loss: 0.0119 - val_acc: 1.0000\n",
      "Epoch 167/200\n",
      " - 0s - loss: 5.2837e-05 - acc: 1.0000 - val_loss: 0.0120 - val_acc: 1.0000\n",
      "Epoch 168/200\n",
      " - 0s - loss: 3.9765e-05 - acc: 1.0000 - val_loss: 0.0120 - val_acc: 1.0000\n",
      "Epoch 169/200\n",
      " - 0s - loss: 1.3206e-04 - acc: 1.0000 - val_loss: 0.0120 - val_acc: 1.0000\n",
      "Epoch 170/200\n",
      " - 0s - loss: 4.5184e-05 - acc: 1.0000 - val_loss: 0.0120 - val_acc: 1.0000\n",
      "Epoch 171/200\n",
      " - 0s - loss: 3.8938e-05 - acc: 1.0000 - val_loss: 0.0120 - val_acc: 1.0000\n",
      "Epoch 172/200\n",
      " - 0s - loss: 3.8039e-05 - acc: 1.0000 - val_loss: 0.0119 - val_acc: 1.0000\n",
      "Epoch 173/200\n",
      " - 0s - loss: 7.7546e-05 - acc: 1.0000 - val_loss: 0.0118 - val_acc: 1.0000\n",
      "Epoch 174/200\n",
      " - 0s - loss: 3.7317e-05 - acc: 1.0000 - val_loss: 0.0117 - val_acc: 1.0000\n",
      "Epoch 175/200\n",
      " - 0s - loss: 4.6776e-05 - acc: 1.0000 - val_loss: 0.0116 - val_acc: 1.0000\n",
      "Epoch 176/200\n",
      " - 0s - loss: 5.1033e-05 - acc: 1.0000 - val_loss: 0.0115 - val_acc: 1.0000\n",
      "Epoch 177/200\n",
      " - 0s - loss: 3.6686e-05 - acc: 1.0000 - val_loss: 0.0115 - val_acc: 1.0000\n",
      "Epoch 178/200\n",
      " - 0s - loss: 5.7414e-05 - acc: 1.0000 - val_loss: 0.0114 - val_acc: 1.0000\n",
      "Epoch 179/200\n",
      " - 0s - loss: 3.6357e-05 - acc: 1.0000 - val_loss: 0.0114 - val_acc: 1.0000\n",
      "Epoch 180/200\n",
      " - 0s - loss: 3.5129e-05 - acc: 1.0000 - val_loss: 0.0114 - val_acc: 1.0000\n",
      "Epoch 181/200\n",
      " - 0s - loss: 3.3520e-05 - acc: 1.0000 - val_loss: 0.0114 - val_acc: 1.0000\n",
      "Epoch 182/200\n",
      " - 0s - loss: 3.9810e-05 - acc: 1.0000 - val_loss: 0.0113 - val_acc: 1.0000\n",
      "Epoch 183/200\n",
      " - 0s - loss: 3.3680e-05 - acc: 1.0000 - val_loss: 0.0113 - val_acc: 1.0000\n",
      "Epoch 184/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " - 0s - loss: 3.4069e-05 - acc: 1.0000 - val_loss: 0.0113 - val_acc: 1.0000\n",
      "Epoch 185/200\n",
      " - 0s - loss: 3.5030e-05 - acc: 1.0000 - val_loss: 0.0112 - val_acc: 1.0000\n",
      "Epoch 186/200\n",
      " - 0s - loss: 4.1208e-05 - acc: 1.0000 - val_loss: 0.0111 - val_acc: 1.0000\n",
      "Epoch 187/200\n",
      " - 0s - loss: 3.3469e-05 - acc: 1.0000 - val_loss: 0.0111 - val_acc: 1.0000\n",
      "Epoch 188/200\n",
      " - 0s - loss: 3.6846e-05 - acc: 1.0000 - val_loss: 0.0111 - val_acc: 1.0000\n",
      "Epoch 189/200\n",
      " - 0s - loss: 4.0005e-05 - acc: 1.0000 - val_loss: 0.0111 - val_acc: 1.0000\n",
      "Epoch 190/200\n",
      " - 0s - loss: 3.2380e-05 - acc: 1.0000 - val_loss: 0.0111 - val_acc: 1.0000\n",
      "Epoch 191/200\n",
      " - 0s - loss: 3.1353e-05 - acc: 1.0000 - val_loss: 0.0111 - val_acc: 1.0000\n",
      "Epoch 192/200\n",
      " - 0s - loss: 3.3316e-05 - acc: 1.0000 - val_loss: 0.0111 - val_acc: 1.0000\n",
      "Epoch 193/200\n",
      " - 0s - loss: 3.6925e-05 - acc: 1.0000 - val_loss: 0.0110 - val_acc: 1.0000\n",
      "Epoch 194/200\n",
      " - 0s - loss: 3.6457e-05 - acc: 1.0000 - val_loss: 0.0111 - val_acc: 1.0000\n",
      "Epoch 195/200\n",
      " - 0s - loss: 3.6666e-05 - acc: 1.0000 - val_loss: 0.0110 - val_acc: 1.0000\n",
      "Epoch 196/200\n",
      " - 0s - loss: 3.2298e-05 - acc: 1.0000 - val_loss: 0.0110 - val_acc: 1.0000\n",
      "Epoch 197/200\n",
      " - 0s - loss: 3.0983e-05 - acc: 1.0000 - val_loss: 0.0109 - val_acc: 1.0000\n",
      "Epoch 198/200\n",
      " - 0s - loss: 3.4055e-05 - acc: 1.0000 - val_loss: 0.0110 - val_acc: 1.0000\n",
      "Epoch 199/200\n",
      " - 0s - loss: 3.2368e-05 - acc: 1.0000 - val_loss: 0.0110 - val_acc: 1.0000\n",
      "Epoch 200/200\n",
      " - 0s - loss: 4.3762e-05 - acc: 1.0000 - val_loss: 0.0109 - val_acc: 1.0000\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f431f95f7b8>"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(x_train, y_train, \n",
    "          batch_size=100, \n",
    "          epochs=200, \n",
    "          validation_data=(x_valid, y_valid), \n",
    "          verbose=2, \n",
    "          callbacks=[])\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## use the network to predict the activity of the ligands in the test set"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "we now run our network on the test set, and compare its predictions "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100.0% of the active ligands were correctly identified\n",
      " 99.0% of the inactive ligands were correctly identified\n"
     ]
    }
   ],
   "source": [
    "# Predict test dataset\n",
    "y_test_predicted = model.predict(x_test)\n",
    "\n",
    "# plot the results\n",
    "for gr in [0,1]:\n",
    "    data_gr = y_test_predicted[y_test_wanted==gr,:]\n",
    "    plt.plot(data_gr[:,0],data_gr[:,1],marker='o',linewidth=0)\n",
    "    \n",
    "plt.xlabel('output bit ACTIVE')\n",
    "plt.ylabel('output bit INACTIVE')\n",
    "plt.plot([0,1],[0,1],'--',color='red')\n",
    "plt.show()\n",
    "\n",
    "score_1 = ((y_test_predicted[:,1]>y_test_predicted[:,0])[y_test_wanted==1]).mean()\n",
    "score_0 = ((y_test_predicted[:,1]<y_test_predicted[:,0])[y_test_wanted==0]).mean()\n",
    "print(\"%5.1f%% of the active ligands were correctly identified\"%(100*score_1))\n",
    "print(\"%5.1f%% of the inactive ligands were correctly identified\"%(100*score_0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "## fit a second model with normalization and dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "batch_normalization_20 (Batc (None, 1000)              4000      \n",
      "_________________________________________________________________\n",
      "dense_37 (Dense)             (None, 32)                32032     \n",
      "_________________________________________________________________\n",
      "dropout_17 (Dropout)         (None, 32)                0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_21 (Batc (None, 32)                128       \n",
      "_________________________________________________________________\n",
      "dense_38 (Dense)             (None, 16)                528       \n",
      "_________________________________________________________________\n",
      "dropout_18 (Dropout)         (None, 16)                0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_22 (Batc (None, 16)                64        \n",
      "_________________________________________________________________\n",
      "dense_39 (Dense)             (None, 16)                272       \n",
      "_________________________________________________________________\n",
      "dropout_19 (Dropout)         (None, 16)                0         \n",
      "_________________________________________________________________\n",
      "dense_40 (Dense)             (None, 2)                 34        \n",
      "=================================================================\n",
      "Total params: 37,058\n",
      "Trainable params: 34,962\n",
      "Non-trainable params: 2,096\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model2 = keras.models.Sequential([\n",
    "    keras.layers.normalization.BatchNormalization(input_shape=tuple([x_train.shape[1]])),\n",
    "    keras.layers.core.Dense(32, activation='relu'),\n",
    "    keras.layers.core.Dropout(rate=0.5),\n",
    "    keras.layers.normalization.BatchNormalization(),\n",
    "    keras.layers.core.Dense(16, activation='relu'),\n",
    "    keras.layers.core.Dropout(rate=0.5),\n",
    "    keras.layers.normalization.BatchNormalization(),\n",
    "    keras.layers.core.Dense(16, activation='relu'),\n",
    "    keras.layers.core.Dropout(rate=0.5),\n",
    "    keras.layers.core.Dense(2,   activation='sigmoid')\n",
    "])\n",
    "\n",
    "model2.compile(loss=\"categorical_crossentropy\", optimizer=\"adadelta\",metrics=[\"accuracy\"])\n",
    "print(model2.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 140 samples, validate on 60 samples\n",
      "Epoch 1/200\n",
      " - 1s - loss: 0.8584 - acc: 0.4500 - val_loss: 0.7144 - val_acc: 0.3500\n",
      "Epoch 2/200\n",
      " - 0s - loss: 0.7700 - acc: 0.5357 - val_loss: 0.7101 - val_acc: 0.5000\n",
      "Epoch 3/200\n",
      " - 0s - loss: 0.7749 - acc: 0.5643 - val_loss: 0.7006 - val_acc: 0.4833\n",
      "Epoch 4/200\n",
      " - 0s - loss: 0.8096 - acc: 0.4857 - val_loss: 0.6880 - val_acc: 0.5000\n",
      "Epoch 5/200\n",
      " - 0s - loss: 0.7554 - acc: 0.4857 - val_loss: 0.6800 - val_acc: 0.5167\n",
      "Epoch 6/200\n",
      " - 0s - loss: 0.7020 - acc: 0.6143 - val_loss: 0.6756 - val_acc: 0.5500\n",
      "Epoch 7/200\n",
      " - 0s - loss: 0.7442 - acc: 0.5000 - val_loss: 0.6637 - val_acc: 0.5833\n",
      "Epoch 8/200\n",
      " - 0s - loss: 0.7567 - acc: 0.5286 - val_loss: 0.6601 - val_acc: 0.6000\n",
      "Epoch 9/200\n",
      " - 0s - loss: 0.7524 - acc: 0.4786 - val_loss: 0.6558 - val_acc: 0.6167\n",
      "Epoch 10/200\n",
      " - 0s - loss: 0.7132 - acc: 0.5857 - val_loss: 0.6451 - val_acc: 0.6833\n",
      "Epoch 11/200\n",
      " - 0s - loss: 0.7015 - acc: 0.5643 - val_loss: 0.6403 - val_acc: 0.7667\n",
      "Epoch 12/200\n",
      " - 0s - loss: 0.6370 - acc: 0.6500 - val_loss: 0.6378 - val_acc: 0.7500\n",
      "Epoch 13/200\n",
      " - 0s - loss: 0.6701 - acc: 0.6071 - val_loss: 0.6326 - val_acc: 0.7333\n",
      "Epoch 14/200\n",
      " - 0s - loss: 0.6992 - acc: 0.5929 - val_loss: 0.6300 - val_acc: 0.7500\n",
      "Epoch 15/200\n",
      " - 0s - loss: 0.6971 - acc: 0.5500 - val_loss: 0.6229 - val_acc: 0.7833\n",
      "Epoch 16/200\n",
      " - 0s - loss: 0.6621 - acc: 0.5786 - val_loss: 0.6168 - val_acc: 0.8167\n",
      "Epoch 17/200\n",
      " - 0s - loss: 0.6655 - acc: 0.6143 - val_loss: 0.6143 - val_acc: 0.7833\n",
      "Epoch 18/200\n",
      " - 0s - loss: 0.6385 - acc: 0.5929 - val_loss: 0.6102 - val_acc: 0.8167\n",
      "Epoch 19/200\n",
      " - 0s - loss: 0.6283 - acc: 0.5929 - val_loss: 0.6051 - val_acc: 0.8500\n",
      "Epoch 20/200\n",
      " - 0s - loss: 0.6682 - acc: 0.6000 - val_loss: 0.6033 - val_acc: 0.8500\n",
      "Epoch 21/200\n",
      " - 0s - loss: 0.6518 - acc: 0.6143 - val_loss: 0.5991 - val_acc: 0.8500\n",
      "Epoch 22/200\n",
      " - 0s - loss: 0.6173 - acc: 0.5643 - val_loss: 0.5927 - val_acc: 0.8500\n",
      "Epoch 23/200\n",
      " - 0s - loss: 0.6536 - acc: 0.6214 - val_loss: 0.5888 - val_acc: 0.8333\n",
      "Epoch 24/200\n",
      " - 0s - loss: 0.6198 - acc: 0.6071 - val_loss: 0.5844 - val_acc: 0.8333\n",
      "Epoch 25/200\n",
      " - 0s - loss: 0.5980 - acc: 0.6786 - val_loss: 0.5784 - val_acc: 0.8167\n",
      "Epoch 26/200\n",
      " - 0s - loss: 0.6091 - acc: 0.6857 - val_loss: 0.5727 - val_acc: 0.8000\n",
      "Epoch 27/200\n",
      " - 0s - loss: 0.6332 - acc: 0.6071 - val_loss: 0.5673 - val_acc: 0.8167\n",
      "Epoch 28/200\n",
      " - 0s - loss: 0.5857 - acc: 0.6643 - val_loss: 0.5629 - val_acc: 0.8167\n",
      "Epoch 29/200\n",
      " - 0s - loss: 0.6052 - acc: 0.6214 - val_loss: 0.5579 - val_acc: 0.8167\n",
      "Epoch 30/200\n",
      " - 0s - loss: 0.6260 - acc: 0.6071 - val_loss: 0.5535 - val_acc: 0.8500\n",
      "Epoch 31/200\n",
      " - 0s - loss: 0.5995 - acc: 0.6786 - val_loss: 0.5494 - val_acc: 0.8333\n",
      "Epoch 32/200\n",
      " - 0s - loss: 0.6008 - acc: 0.6143 - val_loss: 0.5443 - val_acc: 0.8333\n",
      "Epoch 33/200\n",
      " - 0s - loss: 0.5590 - acc: 0.7429 - val_loss: 0.5402 - val_acc: 0.8667\n",
      "Epoch 34/200\n",
      " - 0s - loss: 0.5514 - acc: 0.6786 - val_loss: 0.5368 - val_acc: 0.8833\n",
      "Epoch 35/200\n",
      " - 0s - loss: 0.5348 - acc: 0.7071 - val_loss: 0.5343 - val_acc: 0.8667\n",
      "Epoch 36/200\n",
      " - 0s - loss: 0.5378 - acc: 0.6786 - val_loss: 0.5313 - val_acc: 0.8500\n",
      "Epoch 37/200\n",
      " - 0s - loss: 0.5278 - acc: 0.7500 - val_loss: 0.5271 - val_acc: 0.8500\n",
      "Epoch 38/200\n",
      " - 0s - loss: 0.5669 - acc: 0.7500 - val_loss: 0.5205 - val_acc: 0.8667\n",
      "Epoch 39/200\n",
      " - 0s - loss: 0.5680 - acc: 0.7357 - val_loss: 0.5147 - val_acc: 0.8667\n",
      "Epoch 40/200\n",
      " - 0s - loss: 0.5565 - acc: 0.7214 - val_loss: 0.5105 - val_acc: 0.8667\n",
      "Epoch 41/200\n",
      " - 0s - loss: 0.5229 - acc: 0.7500 - val_loss: 0.5046 - val_acc: 0.8667\n",
      "Epoch 42/200\n",
      " - 0s - loss: 0.5327 - acc: 0.6929 - val_loss: 0.5000 - val_acc: 0.8667\n",
      "Epoch 43/200\n",
      " - 0s - loss: 0.5037 - acc: 0.7643 - val_loss: 0.4928 - val_acc: 0.8667\n",
      "Epoch 44/200\n",
      " - 0s - loss: 0.4984 - acc: 0.7857 - val_loss: 0.4878 - val_acc: 0.8667\n",
      "Epoch 45/200\n",
      " - 0s - loss: 0.5084 - acc: 0.7429 - val_loss: 0.4840 - val_acc: 0.8667\n",
      "Epoch 46/200\n",
      " - 0s - loss: 0.5142 - acc: 0.7429 - val_loss: 0.4791 - val_acc: 0.8667\n",
      "Epoch 47/200\n",
      " - 0s - loss: 0.5065 - acc: 0.7571 - val_loss: 0.4760 - val_acc: 0.8667\n",
      "Epoch 48/200\n",
      " - 0s - loss: 0.4981 - acc: 0.7357 - val_loss: 0.4701 - val_acc: 0.8667\n",
      "Epoch 49/200\n",
      " - 0s - loss: 0.4633 - acc: 0.8000 - val_loss: 0.4634 - val_acc: 0.8833\n",
      "Epoch 50/200\n",
      " - 0s - loss: 0.4926 - acc: 0.7214 - val_loss: 0.4581 - val_acc: 0.8833\n",
      "Epoch 51/200\n",
      " - 0s - loss: 0.4848 - acc: 0.7500 - val_loss: 0.4510 - val_acc: 0.9167\n",
      "Epoch 52/200\n",
      " - 0s - loss: 0.4695 - acc: 0.7643 - val_loss: 0.4441 - val_acc: 0.9000\n",
      "Epoch 53/200\n",
      " - 0s - loss: 0.5527 - acc: 0.7000 - val_loss: 0.4363 - val_acc: 0.9167\n",
      "Epoch 54/200\n",
      " - 0s - loss: 0.4412 - acc: 0.7857 - val_loss: 0.4340 - val_acc: 0.9167\n",
      "Epoch 55/200\n",
      " - 0s - loss: 0.4523 - acc: 0.7786 - val_loss: 0.4285 - val_acc: 0.9167\n",
      "Epoch 56/200\n",
      " - 0s - loss: 0.4532 - acc: 0.8071 - val_loss: 0.4257 - val_acc: 0.9333\n",
      "Epoch 57/200\n",
      " - 0s - loss: 0.4650 - acc: 0.7714 - val_loss: 0.4193 - val_acc: 0.9167\n",
      "Epoch 58/200\n",
      " - 0s - loss: 0.4436 - acc: 0.8214 - val_loss: 0.4143 - val_acc: 0.9167\n",
      "Epoch 59/200\n",
      " - 0s - loss: 0.4825 - acc: 0.8429 - val_loss: 0.4083 - val_acc: 0.9167\n",
      "Epoch 60/200\n",
      " - 0s - loss: 0.4893 - acc: 0.7357 - val_loss: 0.4026 - val_acc: 0.8833\n",
      "Epoch 61/200\n",
      " - 0s - loss: 0.4315 - acc: 0.8357 - val_loss: 0.3943 - val_acc: 0.9167\n",
      "Epoch 62/200\n",
      " - 0s - loss: 0.4077 - acc: 0.8357 - val_loss: 0.3907 - val_acc: 0.9167\n",
      "Epoch 63/200\n",
      " - 0s - loss: 0.4351 - acc: 0.8071 - val_loss: 0.3839 - val_acc: 0.9167\n",
      "Epoch 64/200\n",
      " - 0s - loss: 0.4530 - acc: 0.8357 - val_loss: 0.3768 - val_acc: 0.9333\n",
      "Epoch 65/200\n",
      " - 0s - loss: 0.3570 - acc: 0.8500 - val_loss: 0.3685 - val_acc: 0.9667\n",
      "Epoch 66/200\n",
      " - 0s - loss: 0.3840 - acc: 0.8000 - val_loss: 0.3634 - val_acc: 0.9833\n",
      "Epoch 67/200\n",
      " - 0s - loss: 0.3806 - acc: 0.8786 - val_loss: 0.3566 - val_acc: 0.9667\n",
      "Epoch 68/200\n",
      " - 0s - loss: 0.4290 - acc: 0.8357 - val_loss: 0.3510 - val_acc: 0.9833\n",
      "Epoch 69/200\n",
      " - 0s - loss: 0.3907 - acc: 0.8929 - val_loss: 0.3459 - val_acc: 0.9667\n",
      "Epoch 70/200\n",
      " - 0s - loss: 0.3804 - acc: 0.8500 - val_loss: 0.3386 - val_acc: 0.9833\n",
      "Epoch 71/200\n",
      " - 0s - loss: 0.3769 - acc: 0.8786 - val_loss: 0.3320 - val_acc: 0.9833\n",
      "Epoch 72/200\n",
      " - 0s - loss: 0.3615 - acc: 0.8500 - val_loss: 0.3258 - val_acc: 1.0000\n",
      "Epoch 73/200\n",
      " - 0s - loss: 0.3759 - acc: 0.8357 - val_loss: 0.3201 - val_acc: 0.9833\n",
      "Epoch 74/200\n",
      " - 0s - loss: 0.3784 - acc: 0.8000 - val_loss: 0.3146 - val_acc: 0.9833\n",
      "Epoch 75/200\n",
      " - 0s - loss: 0.3799 - acc: 0.8643 - val_loss: 0.3107 - val_acc: 1.0000\n",
      "Epoch 76/200\n",
      " - 0s - loss: 0.3805 - acc: 0.8429 - val_loss: 0.3051 - val_acc: 1.0000\n",
      "Epoch 77/200\n",
      " - 0s - loss: 0.3636 - acc: 0.8714 - val_loss: 0.3006 - val_acc: 1.0000\n",
      "Epoch 78/200\n",
      " - 0s - loss: 0.3385 - acc: 0.8786 - val_loss: 0.2938 - val_acc: 1.0000\n",
      "Epoch 79/200\n",
      " - 0s - loss: 0.3751 - acc: 0.8714 - val_loss: 0.2889 - val_acc: 1.0000\n",
      "Epoch 80/200\n",
      " - 0s - loss: 0.2909 - acc: 0.8786 - val_loss: 0.2837 - val_acc: 1.0000\n",
      "Epoch 81/200\n",
      " - 0s - loss: 0.3843 - acc: 0.8786 - val_loss: 0.2792 - val_acc: 1.0000\n",
      "Epoch 82/200\n",
      " - 0s - loss: 0.3236 - acc: 0.8643 - val_loss: 0.2731 - val_acc: 1.0000\n",
      "Epoch 83/200\n",
      " - 0s - loss: 0.3194 - acc: 0.8929 - val_loss: 0.2728 - val_acc: 1.0000\n",
      "Epoch 84/200\n",
      " - 0s - loss: 0.3025 - acc: 0.9071 - val_loss: 0.2662 - val_acc: 1.0000\n",
      "Epoch 85/200\n",
      " - 0s - loss: 0.3856 - acc: 0.8286 - val_loss: 0.2630 - val_acc: 1.0000\n",
      "Epoch 86/200\n",
      " - 0s - loss: 0.2797 - acc: 0.9286 - val_loss: 0.2576 - val_acc: 1.0000\n",
      "Epoch 87/200\n",
      " - 0s - loss: 0.3356 - acc: 0.8643 - val_loss: 0.2539 - val_acc: 1.0000\n",
      "Epoch 88/200\n",
      " - 0s - loss: 0.3407 - acc: 0.8643 - val_loss: 0.2474 - val_acc: 1.0000\n",
      "Epoch 89/200\n",
      " - 0s - loss: 0.3130 - acc: 0.8714 - val_loss: 0.2459 - val_acc: 1.0000\n",
      "Epoch 90/200\n",
      " - 0s - loss: 0.2838 - acc: 0.9143 - val_loss: 0.2423 - val_acc: 1.0000\n",
      "Epoch 91/200\n",
      " - 0s - loss: 0.2936 - acc: 0.9071 - val_loss: 0.2380 - val_acc: 1.0000\n",
      "Epoch 92/200\n",
      " - 0s - loss: 0.2614 - acc: 0.9214 - val_loss: 0.2323 - val_acc: 1.0000\n",
      "Epoch 93/200\n",
      " - 0s - loss: 0.2395 - acc: 0.9071 - val_loss: 0.2280 - val_acc: 1.0000\n",
      "Epoch 94/200\n",
      " - 0s - loss: 0.2330 - acc: 0.9500 - val_loss: 0.2200 - val_acc: 1.0000\n",
      "Epoch 95/200\n",
      " - 0s - loss: 0.2984 - acc: 0.9071 - val_loss: 0.2156 - val_acc: 1.0000\n",
      "Epoch 96/200\n",
      " - 0s - loss: 0.3001 - acc: 0.8929 - val_loss: 0.2099 - val_acc: 1.0000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 97/200\n",
      " - 0s - loss: 0.2563 - acc: 0.9143 - val_loss: 0.2057 - val_acc: 1.0000\n",
      "Epoch 98/200\n",
      " - 0s - loss: 0.2926 - acc: 0.8786 - val_loss: 0.2012 - val_acc: 1.0000\n",
      "Epoch 99/200\n",
      " - 0s - loss: 0.2661 - acc: 0.8786 - val_loss: 0.1987 - val_acc: 1.0000\n",
      "Epoch 100/200\n",
      " - 0s - loss: 0.2747 - acc: 0.8714 - val_loss: 0.1941 - val_acc: 1.0000\n",
      "Epoch 101/200\n",
      " - 0s - loss: 0.3049 - acc: 0.8929 - val_loss: 0.1906 - val_acc: 1.0000\n",
      "Epoch 102/200\n",
      " - 0s - loss: 0.3018 - acc: 0.8786 - val_loss: 0.1867 - val_acc: 1.0000\n",
      "Epoch 103/200\n",
      " - 0s - loss: 0.2845 - acc: 0.9000 - val_loss: 0.1832 - val_acc: 1.0000\n",
      "Epoch 104/200\n",
      " - 0s - loss: 0.2659 - acc: 0.9214 - val_loss: 0.1796 - val_acc: 1.0000\n",
      "Epoch 105/200\n",
      " - 0s - loss: 0.2589 - acc: 0.8857 - val_loss: 0.1757 - val_acc: 1.0000\n",
      "Epoch 106/200\n",
      " - 0s - loss: 0.2885 - acc: 0.8857 - val_loss: 0.1723 - val_acc: 1.0000\n",
      "Epoch 107/200\n",
      " - 0s - loss: 0.2107 - acc: 0.9429 - val_loss: 0.1679 - val_acc: 1.0000\n",
      "Epoch 108/200\n",
      " - 0s - loss: 0.2715 - acc: 0.9071 - val_loss: 0.1636 - val_acc: 1.0000\n",
      "Epoch 109/200\n",
      " - 0s - loss: 0.2685 - acc: 0.9000 - val_loss: 0.1596 - val_acc: 1.0000\n",
      "Epoch 110/200\n",
      " - 0s - loss: 0.2243 - acc: 0.9143 - val_loss: 0.1560 - val_acc: 1.0000\n",
      "Epoch 111/200\n",
      " - 0s - loss: 0.2068 - acc: 0.9571 - val_loss: 0.1514 - val_acc: 1.0000\n",
      "Epoch 112/200\n",
      " - 0s - loss: 0.2226 - acc: 0.9286 - val_loss: 0.1491 - val_acc: 1.0000\n",
      "Epoch 113/200\n",
      " - 0s - loss: 0.2317 - acc: 0.9286 - val_loss: 0.1461 - val_acc: 1.0000\n",
      "Epoch 114/200\n",
      " - 0s - loss: 0.2300 - acc: 0.9286 - val_loss: 0.1422 - val_acc: 1.0000\n",
      "Epoch 115/200\n",
      " - 0s - loss: 0.2285 - acc: 0.9286 - val_loss: 0.1380 - val_acc: 1.0000\n",
      "Epoch 116/200\n",
      " - 0s - loss: 0.2926 - acc: 0.8786 - val_loss: 0.1352 - val_acc: 1.0000\n",
      "Epoch 117/200\n",
      " - 0s - loss: 0.2988 - acc: 0.9071 - val_loss: 0.1330 - val_acc: 1.0000\n",
      "Epoch 118/200\n",
      " - 0s - loss: 0.2234 - acc: 0.9357 - val_loss: 0.1291 - val_acc: 1.0000\n",
      "Epoch 119/200\n",
      " - 0s - loss: 0.1856 - acc: 0.9571 - val_loss: 0.1259 - val_acc: 1.0000\n",
      "Epoch 120/200\n",
      " - 0s - loss: 0.2235 - acc: 0.9286 - val_loss: 0.1243 - val_acc: 1.0000\n",
      "Epoch 121/200\n",
      " - 0s - loss: 0.2051 - acc: 0.9571 - val_loss: 0.1190 - val_acc: 1.0000\n",
      "Epoch 122/200\n",
      " - 0s - loss: 0.2517 - acc: 0.9286 - val_loss: 0.1169 - val_acc: 1.0000\n",
      "Epoch 123/200\n",
      " - 0s - loss: 0.2147 - acc: 0.9214 - val_loss: 0.1145 - val_acc: 1.0000\n",
      "Epoch 124/200\n",
      " - 0s - loss: 0.2255 - acc: 0.9214 - val_loss: 0.1120 - val_acc: 1.0000\n",
      "Epoch 125/200\n",
      " - 0s - loss: 0.2488 - acc: 0.8786 - val_loss: 0.1097 - val_acc: 1.0000\n",
      "Epoch 126/200\n",
      " - 0s - loss: 0.1807 - acc: 0.9357 - val_loss: 0.1066 - val_acc: 1.0000\n",
      "Epoch 127/200\n",
      " - 0s - loss: 0.2061 - acc: 0.9143 - val_loss: 0.1048 - val_acc: 1.0000\n",
      "Epoch 128/200\n",
      " - 0s - loss: 0.2157 - acc: 0.9429 - val_loss: 0.1024 - val_acc: 1.0000\n",
      "Epoch 129/200\n",
      " - 0s - loss: 0.2079 - acc: 0.9286 - val_loss: 0.1001 - val_acc: 1.0000\n",
      "Epoch 130/200\n",
      " - 0s - loss: 0.1920 - acc: 0.9286 - val_loss: 0.0977 - val_acc: 1.0000\n",
      "Epoch 131/200\n",
      " - 0s - loss: 0.2163 - acc: 0.9429 - val_loss: 0.0943 - val_acc: 1.0000\n",
      "Epoch 132/200\n",
      " - 0s - loss: 0.1468 - acc: 0.9786 - val_loss: 0.0922 - val_acc: 1.0000\n",
      "Epoch 133/200\n",
      " - 0s - loss: 0.1639 - acc: 0.9500 - val_loss: 0.0906 - val_acc: 1.0000\n",
      "Epoch 134/200\n",
      " - 0s - loss: 0.2001 - acc: 0.9286 - val_loss: 0.0882 - val_acc: 1.0000\n",
      "Epoch 135/200\n",
      " - 0s - loss: 0.2091 - acc: 0.9214 - val_loss: 0.0859 - val_acc: 1.0000\n",
      "Epoch 136/200\n",
      " - 0s - loss: 0.1808 - acc: 0.9429 - val_loss: 0.0837 - val_acc: 1.0000\n",
      "Epoch 137/200\n",
      " - 0s - loss: 0.1851 - acc: 0.9214 - val_loss: 0.0826 - val_acc: 1.0000\n",
      "Epoch 138/200\n",
      " - 0s - loss: 0.1600 - acc: 0.9500 - val_loss: 0.0814 - val_acc: 1.0000\n",
      "Epoch 139/200\n",
      " - 0s - loss: 0.1527 - acc: 0.9571 - val_loss: 0.0801 - val_acc: 1.0000\n",
      "Epoch 140/200\n",
      " - 0s - loss: 0.1710 - acc: 0.9571 - val_loss: 0.0781 - val_acc: 1.0000\n",
      "Epoch 141/200\n",
      " - 0s - loss: 0.1806 - acc: 0.9357 - val_loss: 0.0763 - val_acc: 1.0000\n",
      "Epoch 142/200\n",
      " - 0s - loss: 0.1779 - acc: 0.9429 - val_loss: 0.0746 - val_acc: 1.0000\n",
      "Epoch 143/200\n",
      " - 0s - loss: 0.1529 - acc: 0.9786 - val_loss: 0.0729 - val_acc: 1.0000\n",
      "Epoch 144/200\n",
      " - 0s - loss: 0.1726 - acc: 0.9714 - val_loss: 0.0718 - val_acc: 1.0000\n",
      "Epoch 145/200\n",
      " - 0s - loss: 0.1868 - acc: 0.9143 - val_loss: 0.0703 - val_acc: 1.0000\n",
      "Epoch 146/200\n",
      " - 0s - loss: 0.1518 - acc: 0.9571 - val_loss: 0.0700 - val_acc: 1.0000\n",
      "Epoch 147/200\n",
      " - 0s - loss: 0.1160 - acc: 0.9714 - val_loss: 0.0684 - val_acc: 1.0000\n",
      "Epoch 148/200\n",
      " - 0s - loss: 0.1265 - acc: 0.9643 - val_loss: 0.0670 - val_acc: 1.0000\n",
      "Epoch 149/200\n",
      " - 0s - loss: 0.1510 - acc: 0.9643 - val_loss: 0.0659 - val_acc: 1.0000\n",
      "Epoch 150/200\n",
      " - 0s - loss: 0.1471 - acc: 0.9714 - val_loss: 0.0650 - val_acc: 1.0000\n",
      "Epoch 151/200\n",
      " - 0s - loss: 0.1560 - acc: 0.9500 - val_loss: 0.0642 - val_acc: 1.0000\n",
      "Epoch 152/200\n",
      " - 0s - loss: 0.1618 - acc: 0.9500 - val_loss: 0.0625 - val_acc: 1.0000\n",
      "Epoch 153/200\n",
      " - 0s - loss: 0.1075 - acc: 0.9571 - val_loss: 0.0617 - val_acc: 1.0000\n",
      "Epoch 154/200\n",
      " - 0s - loss: 0.1561 - acc: 0.9500 - val_loss: 0.0600 - val_acc: 1.0000\n",
      "Epoch 155/200\n",
      " - 0s - loss: 0.1700 - acc: 0.9500 - val_loss: 0.0594 - val_acc: 1.0000\n",
      "Epoch 156/200\n",
      " - 0s - loss: 0.1641 - acc: 0.9500 - val_loss: 0.0578 - val_acc: 1.0000\n",
      "Epoch 157/200\n",
      " - 0s - loss: 0.1723 - acc: 0.9643 - val_loss: 0.0576 - val_acc: 1.0000\n",
      "Epoch 158/200\n",
      " - 0s - loss: 0.1570 - acc: 0.9571 - val_loss: 0.0564 - val_acc: 1.0000\n",
      "Epoch 159/200\n",
      " - 0s - loss: 0.1617 - acc: 0.9571 - val_loss: 0.0558 - val_acc: 1.0000\n",
      "Epoch 160/200\n",
      " - 0s - loss: 0.1391 - acc: 0.9786 - val_loss: 0.0543 - val_acc: 1.0000\n",
      "Epoch 161/200\n",
      " - 0s - loss: 0.1712 - acc: 0.9500 - val_loss: 0.0535 - val_acc: 1.0000\n",
      "Epoch 162/200\n",
      " - 0s - loss: 0.1584 - acc: 0.9571 - val_loss: 0.0534 - val_acc: 1.0000\n",
      "Epoch 163/200\n",
      " - 0s - loss: 0.1409 - acc: 0.9500 - val_loss: 0.0530 - val_acc: 1.0000\n",
      "Epoch 164/200\n",
      " - 0s - loss: 0.1488 - acc: 0.9714 - val_loss: 0.0530 - val_acc: 1.0000\n",
      "Epoch 165/200\n",
      " - 0s - loss: 0.1163 - acc: 0.9643 - val_loss: 0.0515 - val_acc: 1.0000\n",
      "Epoch 166/200\n",
      " - 0s - loss: 0.1927 - acc: 0.9500 - val_loss: 0.0507 - val_acc: 1.0000\n",
      "Epoch 167/200\n",
      " - 0s - loss: 0.1124 - acc: 0.9643 - val_loss: 0.0505 - val_acc: 1.0000\n",
      "Epoch 168/200\n",
      " - 0s - loss: 0.1221 - acc: 0.9643 - val_loss: 0.0495 - val_acc: 1.0000\n",
      "Epoch 169/200\n",
      " - 0s - loss: 0.1344 - acc: 0.9571 - val_loss: 0.0493 - val_acc: 1.0000\n",
      "Epoch 170/200\n",
      " - 0s - loss: 0.1151 - acc: 0.9714 - val_loss: 0.0479 - val_acc: 1.0000\n",
      "Epoch 171/200\n",
      " - 0s - loss: 0.1077 - acc: 0.9857 - val_loss: 0.0458 - val_acc: 1.0000\n",
      "Epoch 172/200\n",
      " - 0s - loss: 0.0951 - acc: 0.9786 - val_loss: 0.0449 - val_acc: 1.0000\n",
      "Epoch 173/200\n",
      " - 0s - loss: 0.1103 - acc: 0.9857 - val_loss: 0.0437 - val_acc: 1.0000\n",
      "Epoch 174/200\n",
      " - 0s - loss: 0.1335 - acc: 0.9500 - val_loss: 0.0425 - val_acc: 1.0000\n",
      "Epoch 175/200\n",
      " - 0s - loss: 0.1031 - acc: 0.9857 - val_loss: 0.0415 - val_acc: 1.0000\n",
      "Epoch 176/200\n",
      " - 0s - loss: 0.1560 - acc: 0.9571 - val_loss: 0.0405 - val_acc: 1.0000\n",
      "Epoch 177/200\n",
      " - 0s - loss: 0.1256 - acc: 0.9571 - val_loss: 0.0404 - val_acc: 1.0000\n",
      "Epoch 178/200\n",
      " - 0s - loss: 0.1419 - acc: 0.9643 - val_loss: 0.0400 - val_acc: 1.0000\n",
      "Epoch 179/200\n",
      " - 0s - loss: 0.1199 - acc: 0.9857 - val_loss: 0.0392 - val_acc: 1.0000\n",
      "Epoch 180/200\n",
      " - 0s - loss: 0.1049 - acc: 0.9786 - val_loss: 0.0384 - val_acc: 1.0000\n",
      "Epoch 181/200\n",
      " - 0s - loss: 0.1323 - acc: 0.9643 - val_loss: 0.0377 - val_acc: 1.0000\n",
      "Epoch 182/200\n",
      " - 0s - loss: 0.1347 - acc: 0.9500 - val_loss: 0.0381 - val_acc: 1.0000\n",
      "Epoch 183/200\n",
      " - 0s - loss: 0.1055 - acc: 0.9643 - val_loss: 0.0377 - val_acc: 1.0000\n",
      "Epoch 184/200\n",
      " - 0s - loss: 0.0938 - acc: 0.9786 - val_loss: 0.0369 - val_acc: 1.0000\n",
      "Epoch 185/200\n",
      " - 0s - loss: 0.1143 - acc: 0.9786 - val_loss: 0.0359 - val_acc: 1.0000\n",
      "Epoch 186/200\n",
      " - 0s - loss: 0.1162 - acc: 0.9714 - val_loss: 0.0353 - val_acc: 1.0000\n",
      "Epoch 187/200\n",
      " - 0s - loss: 0.1006 - acc: 0.9786 - val_loss: 0.0346 - val_acc: 1.0000\n",
      "Epoch 188/200\n",
      " - 0s - loss: 0.1383 - acc: 0.9714 - val_loss: 0.0353 - val_acc: 1.0000\n",
      "Epoch 189/200\n",
      " - 0s - loss: 0.1126 - acc: 0.9857 - val_loss: 0.0352 - val_acc: 1.0000\n",
      "Epoch 190/200\n",
      " - 0s - loss: 0.1253 - acc: 0.9643 - val_loss: 0.0338 - val_acc: 1.0000\n",
      "Epoch 191/200\n",
      " - 0s - loss: 0.1139 - acc: 0.9643 - val_loss: 0.0337 - val_acc: 1.0000\n",
      "Epoch 192/200\n",
      " - 0s - loss: 0.0956 - acc: 0.9786 - val_loss: 0.0326 - val_acc: 1.0000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 193/200\n",
      " - 0s - loss: 0.1254 - acc: 0.9429 - val_loss: 0.0314 - val_acc: 1.0000\n",
      "Epoch 194/200\n",
      " - 0s - loss: 0.1316 - acc: 0.9643 - val_loss: 0.0314 - val_acc: 1.0000\n",
      "Epoch 195/200\n",
      " - 0s - loss: 0.1079 - acc: 0.9857 - val_loss: 0.0301 - val_acc: 1.0000\n",
      "Epoch 196/200\n",
      " - 0s - loss: 0.1201 - acc: 0.9643 - val_loss: 0.0297 - val_acc: 1.0000\n",
      "Epoch 197/200\n",
      " - 0s - loss: 0.0967 - acc: 0.9714 - val_loss: 0.0291 - val_acc: 1.0000\n",
      "Epoch 198/200\n",
      " - 0s - loss: 0.0729 - acc: 0.9714 - val_loss: 0.0284 - val_acc: 1.0000\n",
      "Epoch 199/200\n",
      " - 0s - loss: 0.0964 - acc: 0.9786 - val_loss: 0.0273 - val_acc: 1.0000\n",
      "Epoch 200/200\n",
      " - 0s - loss: 0.1470 - acc: 0.9571 - val_loss: 0.0255 - val_acc: 1.0000\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f4325209ac8>"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "callback_early_stopping = keras.callbacks.EarlyStopping(\n",
    "    monitor='val_loss', \n",
    "    patience=10, \n",
    "    verbose=0, \n",
    "    mode='auto')\n",
    "\n",
    "\n",
    "model2.fit(x_train, y_train, \n",
    "          batch_size=100, \n",
    "          epochs=200, \n",
    "          validation_data=(x_valid, y_valid), \n",
    "          verbose=2, \n",
    "          callbacks=[callback_early_stopping])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 99.0% of the active ligands were correctly identified\n",
      "100.0% of the inactive ligands were correctly identified\n"
     ]
    }
   ],
   "source": [
    "# Predict test dataset\n",
    "y_test_predicted = model2.predict(x_test)\n",
    "\n",
    "# plot the results\n",
    "for gr in [0,1]:\n",
    "    data_gr = y_test_predicted[y_test_wanted==gr,:]\n",
    "    plt.plot(data_gr[:,0],data_gr[:,1],marker='o',linewidth=0)\n",
    "    \n",
    "plt.xlabel('output bit ACTIVE')\n",
    "plt.ylabel('output bit INACTIVE')\n",
    "plt.plot([0,1],[0,1],'--',color='red')\n",
    "plt.show()\n",
    "\n",
    "score_1 = ((y_test_predicted[:,1]>y_test_predicted[:,0])[y_test_wanted==1]).mean()\n",
    "score_0 = ((y_test_predicted[:,1]<y_test_predicted[:,0])[y_test_wanted==0]).mean()\n",
    "print(\"%5.1f%% of the active ligands were correctly identified\"%(100*score_1))\n",
    "print(\"%5.1f%% of the inactive ligands were correctly identified\"%(100*score_0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}