project_gru_keras_tsamp50.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import LSTM, GRU, Dense, Dropout, TimeDistributed, BatchNormalization, Flatten, Activation\n",
    "from keras import regularizers\n",
    "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "from keras.models import load_model\n",
    "import numpy as np\n",
    "import h5py\n",
    "import matplotlib.pyplot as plt\n",
    "from data_utils import get_data\n",
    "from sklearn import preprocessing\n",
    "from functools import reduce\n",
    "from scipy.fftpack import dct, idct\n",
    "from scipy import signal\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train: (2458, 1000, 22) \n",
      "y_train: (2458,) \n",
      "X_val: (0, 1000, 22) \n",
      "y_val: (0,) \n",
      "X_test: (100, 1000, 22) \n",
      "y_test: (100,) \n"
     ]
    }
   ],
   "source": [
    "# Load data from all .mat files, combine them, eliminate EOG signals, shuffle and \n",
    "# seperate training data, validation data and testing data.\n",
    "# Also do mean subtraction on x.\n",
    "\n",
    "data = get_data('project_datasets',num_validation=0, num_test=100, subtract_mean=False, subtract_axis=1,\n",
    "                transpose=True)\n",
    "for k in data.keys():\n",
    "    print('{}: {} '.format(k, data[k].shape))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# substract data from list\n",
    "X_train = data.get('X_train')\n",
    "y_train = data.get('y_train')\n",
    "#X_val = data.get('X_val')\n",
    "#y_val = data.get('y_val')\n",
    "X_test = data.get('X_test')\n",
    "y_test = data.get('y_test')\n",
    "\n",
    "# get data dimension\n",
    "N_train, T_train, C_train = data.get('X_train').shape\n",
    "N_val, T_val, C_val = data.get('X_val').shape\n",
    "N_test, T_test, C_test = data.get('X_test').shape\n",
    "\n",
    "# add dummy zeros for y classification\n",
    "lb = preprocessing.LabelBinarizer()\n",
    "lb.fit([0, 1, 2, 3,])\n",
    "lb.classes_\n",
    "y_train = lb.transform(y_train)\n",
    "if (N_val > 0): y_val = lb.transform(y_val)\n",
    "y_test = lb.transform(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Filtering through FFT(discrete cosine transform)\n",
    "\n",
    "def filter(x, low=0, high=1, plot=False):\n",
    "    N = x.shape[0]\n",
    "    t = np.linspace(0,N,N)\n",
    "    y = dct(x, norm='ortho')\n",
    "    window = np.zeros(N)\n",
    "    window[int(low*N):int(high*N)] = 1\n",
    "    yr = idct(y*window, norm='ortho')\n",
    "    sum(abs(x-yr)**2) / sum(abs(x)**2)\n",
    "    if plot:\n",
    "        plt.plot(t, x, '-b')\n",
    "        plt.plot(t, yr, 'r')\n",
    "    return x\n",
    "\n",
    "# Filter band\n",
    "low_freq = 0.02\n",
    "high_freq = 0.4\n",
    "\n",
    "for i in np.arange(N_train):\n",
    "    for j in np.arange(C_train):\n",
    "       X_train[i,:,j]= filter(X_train[i,:,j],low_freq,high_freq)\n",
    "for i in np.arange(N_val):\n",
    "    for j in arange(C_val):\n",
    "       X_val[i,:,j]= filter(X_val[i,:,j],low_freq,high_freq)\n",
    "    \n",
    "for i in np.arange(N_test):\n",
    "    for j in np.arange(C_test):\n",
    "       X_test[i,:,j]= filter(X_test[i,:,j],low_freq,high_freq)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f01a0297390>]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXwAAAD8CAYAAAB0IB+mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzsnXd4HNX1/j+zu6qWq1xwl22Mq2yDG90G3IDQzJeEkgAhhBLgFwglQAIJgdAJJCEJEEgwoSQQIFR3sI3BuHdsY2y5d1mWZfXVzu+Pu3fnzuzMbNGqz/s8enY1O+XOzL3vvHPOuedouq7jwYMHDx6aP3wN3QAPHjx48FA/8AjfgwcPHloIPML34MGDhxYCj/A9ePDgoYXAI3wPHjx4aCHwCN+DBw8eWgg8wvfgwYOHFgKP8D148OChhcAjfA8ePHhoIQg0dANUdOzYUc/Ly2voZnjw4MFDk8Ly5csP6breKdZ6jYrw8/LyWLZsWUM3w4MHDx6aFDRN2x7Pep5Jx4MHDx5aCDzC9+DBg4cWAo/wPXjw4KGFwCN8Dx48eGgh8AjfgwcPHloIPML34MGDhxYCj/A9ePDgoYXAI3wbhELwz39CdXVDt8SDBw8eUgeP8G0wbRpcdx384Q8N3RIPHjx4SB08wrfBgQPis7CwYdvhwYMHD6mER/g2CIXEp8+7Oh48eGhG8CjNBpLw/f6GbYcHDx48pBIe4dugpkZ8egrfgwcPzQkepdnAI3wPHjw0R3iUZgNJ+J5Jx4MHqKqCd98FXW/olnioLTzCt4Fnw/fgwcBvfgP/938wc2ZDt8RDbeERvg08wvfgwcCOHeLTC1Nu+vAI3wbSpKNpDdsODx4aA+Q48Ew6TR8e4dtAKnz56cFDS4ZH+M0HHuHboKJCfHq5dDx48NCc4BG+DY4dE5+S+JsTXnkFHnywoVvhoSnCU/hNHx7h26C4WHw+9hisW9ewbUk1rr8eHn4YysoauiUemgo8X1bzgUf4FpSWwgcfGP8vXdpwbalL7N/f0C3w4MFDfcMjfAs2bWroFtQPPP+Eh3jhOW2bDzzCt6ClROZ4hO8hXiRD+PPnw1131U17PCQPj/AtaCm27fog/BdegMmT6/44Hhofxo+HZ55p6FZ4sCLQ0A1obGgphF9VVffHuPnmuj+Gh/pDMiYdXfecvo0JtVb4mqb11DTtc03TNmiatl7TtJ+Hl3fQNG22pmmbw5/ta9/cukd5ufn/5tRZV6wwvnsmHQ/xojZjwOtnjQupMOkEgTt1XR8EnAzcomnaYOBeYK6u6/2BueH/Gz2as8L/+9+N795A9JAoklH49fEm6SF+1JrwdV3fq+v6ivD3EmAD0B24CJgWXm0acHFtj1UfaM6EH1AMeN5A9BAvahOl4wmLxoWUOm01TcsDTgQWA110Xd8L4qEAdE7lseoKLYXwvYHoIV54Jp3mg5QRvqZpOcC7wO26rh9NYLsbNE1bpmnasoMHD6aqOUmjOXfQtDTje3M+Tw8Ngx//OPrh4L1JNi6khPA1TUtDkP0buq6/F168X9O0ruHfuwIH7LbVdf0lXddH6bo+qlOnTqloTq0gUyM3RzSUwvcm7DQPuN3HsjJ49dXo5Z6waFxIRZSOBrwCbNB1/Q/KTx8C14S/XwN8YN22MaLr5gWcxycN3Yw6QX3a8J991vjeUiazNVfEY9K57DL75R7hNy6kQuGfBvwIOFvTtFXhv/OAx4GJmqZtBiaG/2/0OHPugzzLHQ3djDqBWsGrrgfiL35hfG/Ob03NBb//PSxc6L6Om8L/9FP75Z5Jp3Gh1hOvdF1fCDhpgHNqu//6RrvDW8liPxohdHzNKg6/oUw6HuE3bug6/PrX4nt1tbmfgKHwb7hBKPl27eLft6fwGxe81AphlJXBIw9W0eboLjKo4jj2Ac3L/mxH+LoOTz0Fdekv9wi/caOkxPielgYbN5p/V0XPtm2J7dsj/MYFj/DDePJJmPbwdnwIhs9jGwAzZsD27Q3YsBTCzoa/ZAnccw9ce23dHdcj/MaNQ4fM///0p+Lz+OPhhBPMvyUqgFqqSWfBAggGG7oV0fAIP4wjR6APBZH/5ff//Afy8kQisKYOu7BM2SmPHKm74yZC+PfcIwq0eKg/HLDEz2Vlic8tW2DzZvNviRJ+S1T4ixfDuHHwwAMN3ZJoeIQfRihkJnyp8CWaQyIwO6etfF2vy0iaRAj/qacaXwlGXW9epj0rZs40/5+Zaf5fNenEcx3UdSork29XU0Vhofh8/HFYv75h22JFiyf82bOhZ09Rx7YvW6kknQN0iiL85gDVpLN3r/j0hXtAXRJaUzfpXHqpcZ0aO955J1qVx4L17U4qfDskSvhFRYm1pTkgI8P4/vbbDdcOOzSRblx3uOMO2LULvvlGKPzt9GYrfZsl4atK7fnnhfpKRuG/8gpcd13869/bJNLmOeP99xu6BfHj+9+HwYMT28Z672ur8NUHvFS7LQkq4VsjnhoaLZ7w1WiVvmxlK33ZRp7JvNNcYB3YwaChXBMh/Ouvh3/+M/71p02LvY6H1CFRZ2FtCd+q4kMhYx8e4TdcO+zgEX6Y8JcsEQq/gD5sI49e7MBHE7dFWGAd2KFQ7TIhlpbWvk1NCclco1AIbr218dlyVVhNbpmZsHSp8b8aaWMnDDp0iN6fFBItkfBVX5lH+I0MkvDbUEwuh9lKXwroQzrVdGVvwzYuxbAO1vXrjWUrViROaHub1+WJiWRCDA8ehL/8BU45JfXtSRXsFP6YMcb/FRXG93jeHkIhY58tMUpHHUcq+TcGeIQf7pDShCMVPkRH6jR1WAf2KaeY1d3f/pbY/o4dq32bGjvU62OthhYP5OBXJzfVFZJ1vFv7RXq6+X+VtBMl/KbusE8G6vX0CL+RoSURvh0hfPml8T1Rs0NLmFSjqttkaiXU5+SbZENrraRsNUOohB+PYq+padmEr44zz6TTyCAHZF+2ArCVvmynN0Czc9zaEcLddxvfEw09bAmv66qqT0bh1wfhnXMO9O0LixYlt721X7gRfqIKvyVmSlXP2SP8RgZJcn0o4AhtOUJ7KslkD12bncKPNfjiIfzhw43vTgo/2UGe6rkAixYlnvvFitoSfn0o/M8+g4ICOOOM5La3PpSsCQM/UbKFx3M+8Sj8YFBEb2kavPVW/G1tCvBMOo0YsnP3oYCt9I0s30ZeiyP8eDKDrlljfHdS+MmS3EsvJbedE049FQYMqN0+moLCry2s/cLpwduP76C4OOb+1HN2Ov9HHjHyN115Zew2NiV4Jp1GDElafdlKAX0iy5sL4ZeWGgScCoWvwknhJ0v4f/97ctvZQRJNbf0MTUHh28EpP70drDNt7fpJRw6yihEM+td9MfennrNTn7POBlZ9JU0dnkmnEaOyEjRC5LEtivCbQyx+Tg5cfrn4nmrCt1P4oRC0aiW+q8nakt1fsjh8ODX7WbXK+N6UCP+SS+Jbb9Mms8kGYNas6PXu4FlyKKXzmjkx96mes5PCtxJ8Y8wsmSzUcdbY6mm0eMLv0QOOYx9ZVJhMOgX0IY0g3dndgK1LDd59V3zKjvi739mvlwqFry47/fTa7y9ZWDNAJosNG4zvTcmkE++1tOa+h+jKV+05zK08TynZtNm3OeYEjJZO+KpJp7El3WvxhJ+TA/39Ihonrb9Z4UPzCs2UnS8vz/73VCh8ddmll0J+Powdm/z+koUkZ2tMeaJQ1VpTUvjxIh4Feht/pg0l3MafxYIFC1zXV++j01ulNYtmY79OiUA953gDGNatg3376qY9Klo84QeDcEKaIPwj7ZsX4VvVhex8TqaWVBN+KAS9esVP5KlM1ZAqZa3uJ5k4/MbutI11z1tzlNt5jve5mNe4mqqMHJg/33WbZBR+cwrxTYbwR42C556rm/aoaPGEX10NvWpEDH5Jbh4gBsEOehFCa9Kx+Ha5c8DZkZRqk84FF4iHSzyDWddTq3Ak0dT2lbq2M21V8muMKjZW2ODP+CvtOcLv+RU1BNjV+7SYhK/e75Zu0omX8NX8Q3WJFk/4wSD0Chawm25ktM2kpEREnt3/2wz20C2i8HNyGradycA6iEIh0NDxa/a9MNUKv0sXYVKJd3ZmKpFKwpdmodra8BtjJIqbSSebUu7kGaYzheWMAmBL93Eil7hLEeR4onSaM+Eno/BDofqJ2W/xhF9dDWM6buVobl+efFIQe04O/PrXRmhmTg507tzQLU0cVhINheBdLuXkJ6cC0Uy4c2di+7dT+Cq5Z2QIhS/X+9e/4JZb4t9XbZBKk46MOqqtwm+M1Z/cHvI38BKdOMTDGLX61nccJ7642PFbusJXST7efhgKeQq/XhAMQpfyAgad14eePY3lfj+0PzGP07pvY+rUxm+LtYOdwh/Garou+YBLeTdq/Vdfdd+fVa3EUvg+n9mkc/XV8Ne/2u871TbcVCn8UEicQ1qaO+G/8IL99VP7TWMj/K++cq5IlUEFv231FJ9xFos4NbL8m+xRoiRWmPDtZsnGY8O3XovmZMNP1KQj1/EIvz5QVUW7Y7ugT5+on4acl0f6vp2kacEmSfh2Cj8XkaD8OW6nFdHpLt0GnvUBYmeisCr1eE06jZXwa2rEwz8ry53wb74Zfvzj6OXqNWtMJh1dh9NOM+ZoWHEd/6Bt6V6TugdYvjadhaFTqZoj7PhPPRW9rWfSsf8ea33PpFMP6Fq1HR+6yD5lRV4e1NTQqXJXkyR86yDSaoK0o5h9J06hB7t5kOiAfKf49WAQJk40L7OLqrESt2rScUOqCT9VBBIv4UtYZwvXtUkn2Qea27mkUcW9PM7OXqcyj/Gm31asgJmV40jbsAaKimzNgC09Dj9ZwvcUfj2gR3U4CsdG4ctlnUq3NckOaW1zRqmYfrpv5Pd4heu4g2cZjDknspMdf+fOaLOtXZiiHeE3F4XvFJapXucbbrBvB9QN4ct8NIlg1SrDL2GHH/EverGT5VN+DQiv7qRJcPbZ4vcFnImm6/DFF7bXt6WHZSZq0pF1JTzCrwf0DIYJ30nhIwi/KSp8a5uzyw4BUNU6l1/yBEdpw1/5GaoDd8UK+33ZDWw7hW9V83aE/5OfxN6utlAJvzYZMyXhZ2c7q2JrLhoVda3wX3st8W2WLLFfXlwMfoLcx2MsYyQHTpoS+e3116F1a/F9MWMJBjIcwzPjmXhlFSNNUVA5IVGFL7OcNhmTjqZp/9A07YCmaeuUZR00TZutadrm8Gf7VBwrldB16B3aStCfDt26Ra/QsydoGp2OFTRJwrcOoqwyYb+vap1LIR25j8cYxwKu4o3IOomQYzwmnfT0aDL/xz+ct2vfHjp1ir8NTlDv18iRye9HRk+4mXTcHlbq9agLG35ubuLbOM3DaNMGTmQlx7OFZ7mD7FZGzGanTgbhV5LJ3p5jHSN15IOtM/sZddgmMY/LNs0BiRL+N9+Iz6ak8F8FpliW3QvM1XW9PzA3/H+jQigk0iIXt+ttf7XT06F7d3JLmqbCdyL8yhzBEi9zPYsZw99a3UVbhEx1KsUXr8KXBDcuHL2XlmYuiOEEuV1GRmqKZqj3qzaJ1OKx4bsR/lVXGd/rgtQyMhLfxi2D4whEtrivOZmsLPNvkvABdvQZBytWkBM6GrUPmQnzn/yYP347xTVmX+Kcc2Ku0mSQzMQraEKEr+v6AsA6rC4CpoW/TwMuTsWxUomqKpEWubiDjTlHok8fckuapg3fSlBZ5YbCB9Dx8TP+SuuyA/yOBwHnOrV2HdeN8J9+WnzKNA6xbLR1Sfi13U8swo/X/twUCH84qzlKawroE7XvKMIPhRgb/BIrNm6E8XzOeUwXARFz55p+b+5VsJKZeAVNyKTjgC66ru8FCH82+NSl6mozEVRXC4Vf0tHGYSuRl0f7o4kp/OnTYyYUrBecf775/+ww4Ve27hhZtoKRcPPN3MJfGMFKR4Vv98Czc2JKtSuJXs5SjWWjb66EP2aM8b0uTDrJEL5b2uoRrGINw9DxRRGQOtt8V89TIC2NU6uj7fjlpSGe5B6204tif/uofMuzZyfe5qaEZAm/ySj82kDTtBs0TVumadqyg3G8+tUG6ekweTJMnQqXXQbVh4rJ5TAlnV0Ufl4ebUt2oQXjDyM477zky82lEtaIm5zyQ1SQQTA9O7Ls5puBRx6hkFz+yM8dCd+O1NwUviT6xqLw1bz2ie7H7xftclLo1ofZ3r3wxz+K77m50K6d+N5YFL6jkgyFGMYaVjECiH7AqceqTsuG0aM5NWgQ/vnni3Mdve0dRrOMB3iYRa0mCMJX7Bw//WnibW5KaPYmHQfs1zStK0D40zbCW9f1l3RdH6Xr+qhOqfDWxcDcufD++/Df/0LNdyJCp6yzu8L36SG66zvjCvGTN3jLlhQ0NsXILi+kyJeLHg61u/TS8MzX9u35C7dwJl+gF0fbZMFe4bsRviT6eAlfkmZmZt0QvnSMJQrptHUjfOu5TZ0Kt98O330n2pEdfr42FsJ3NOkUFNCGElYjChdb7696rFAIGDeOEcFlZCNW9PuhTWYVV2+6n9UM4w2uYmHWJNi921RYwDqfo7khmdQK0PRNOh8C14S/XwN8UIfHSg4FgvAruroQfjgWP49tcRFRY3butqoQhC+hJs5agrA99D680nbbZAm/IUw6RUXRoZ+FhcntSyr89PT4CV+mKwgGxblIwq8Lk04yqtCR8FevBogofOv9VWsLVFUBZ55JGkFOYREgzvXayhfpXrGVX/IEmt/PwqwwuytmHbuH/xRryEcTRrImnfpAqsIy3wIWAQM0TduladpPgMeBiZqmbQYmhv9vVNAKRFrkim7uJh0QhB+P47ax3WAV2RWFHPHn0r27+H/oUOO35YjYxYHHltluG69Jx2rDbwiTztat0ctqS/iJmHQkdL3uFX4yE5YcHxKrVlGDj3WIjvH975t/VhV+RQVw2mkE8TMOYdbJqDzKbUd+x1zOZiaTycyEnb7eopJ8mPB1XSTRU9GuXeMWSolCtQQ89FDDtcMOqYrSuULX9a66rqfput5D1/VXdF0v1HX9HF3X+4c/U1RlNHXw7SjgCG1F8LcTevQgpPnIIz7HbWPuuK0qCjni78ipp8KXX4qMoBK/eq4ze/w9GFS+3HZb68PO5xM23lAITjrJiK1vDDZ8uypXR+0tVTGhEr5K7Js2wS9+IQa327nVNeEnM2HN8fquXs0mBlBBFmlp0KGD+ecowm/dmjX+kyKEP3XrU3TUD3EPTwIaGRnh8TBpEsybB5WVLLfpXhkZzXfiVSKoj3KIDe60rS/YDYzAzgK20te9DF5aGsfa9qAP8U2+aswKP6fiEMUBYdI59VSzzfDnP4e93UYxpMKe8K2kJmO0y8th5UrDhOJk0qlPwrezayc7dd9J4X/ve/Dss2KimnXfqqksFBJ+Cagfhf///h9cc439uhKO/XjVqog5R96/Dz6AxYvFd/W6ynP5Km0cY1lMHgVMLfgDb3G5iPwKrx8h/PJy+PJL23uTluYRfn2hxRC+XUdL37VVFCt3CVMDKM7tU2cKf88eeOaZeni66zo5VYc5muY8NXNHp5H0C34r5tgr2LcvOsRTXk+rWcfJaRtLicrfU0H4dkU9kiV81WlbVWXcJxnBEghEn5ssDC5NOoGAePDVhQ3fel4nn2ykhXK6jnbLX3isCHbsiDhspdnnwguN0NIohY8g/AyqeI+p+PVqfs0jkXWys8NEPn686AizZtkSe1VV8yL8ZMeyp/DrEBoh0vdso4A+MQtdH8vNi9uGnyjhX3YZ3HUXfPttYtvFAzlo27QB/Ugxfr3GlfB3HSeqGrHS7Lh97LFoUpOOPyuJyfWSNelkZoqOX5vOb0doqVD4YJyf3F8o5L5vWcnIzQdQG1jvy+WXG/fGqb/aXZ9zOpodtnbpKNRxIu/714HTCaFxIqv4qNtNbKVfZJ2srPC1yckRr5SzZkWuwf/9n7Gvzp2bV/I0z6TTCHEc+/BXVcQ26QDHOubRnd3UlMc2mKo3O54BLp2JqX4NrKkx9hkMwty3xYHW7uvouM2eruFRbjG02jn5ZLZFlfALC8XAHcvX+A/tB5Iz6UB81+Orr2DNmujldtsmm5zNSvjynsr2Vlc7n5tU+LHCOmsDO3OSNNU5iQ+765O1WRC+VPivvx69jqrwy8vFxLujvnasYgRHac20Xubc+VlZykNn0iRYuRJ9v4jOVgm/Tx9lvQMHGrdNJA405ua3WMKXxcnjMekc65gnpojv2BFzv+ogu/VW4cxcu9Z5/brqHJLgNE0MpiPfiUyZhTgr/Mo2ndip9YRl5kgdq4nkvvvg/vvFd3Vyzk3n7eD7/57K15yCNmggTJtGWkA3tccJqsJX/3fDaafB8OHRy+2ILlUK347w3c5N3b4uTDp2x06G8LM3rYIuXdjPcQD06BG9jkr4H30kHvplZXATL3AhH1IUMM+jyc5Wrns4+D7n6zmR3yQCgTDhr14tDvzyy/YNbyLwTDqNEJLw41H4peGJWdr2bTH3qw6mRYuEM3PYsNjrpzo9sNyfHHQycZob4QcCsEIbFaXwrYT/6KPCTASC8NOo4m6eZNrSQfTfOpNH034jYj6vvZZh951PD3bGrfAlWc2ZE9dp2qK2Jp2qKuGQDQbjI3ynPPm6bph0MjPrhvDLyoTDXUUswrdbnrlplf3TU4GdH6yyEpYyhvmMj7ruEZMOiFCuDh1ot3RW5DeJQACC1bpQSNXVwlMcxl//Wrtspw0Bz6TTCNEXEay9nd4xCb+8Sx4Avh3bYu5XHUzxzJyTnSPVZCAJv1Ur0ZEy4yD8tDRYxkiR7lBx3NqZdORbUcbX81nJiTzJL1mQMYlHf/gNT7X6rciV/qc/0WbVfNYzhK4fvYRd4XQJSQyXXSY+nZK42cE6wGqr8J95RoRcvvii4bS1kqhqy3cK+VRNOq1bO2ciTRahkOg3MnWDhJsN/9ix6OuVRhUZW76BESNcj2dXNEUlqaj6C2Gx8eabiAs4YQIdV8wC9MibnGzveUfehIULoXdv+PzzyIC45RbnGg2NFfL6/vjHyc2Erku0WMLvQwFl7btRSWZMk05lx+4E8cdF+IlOq5brO9l3P/gAPvss9n6sUBU+QHpJfAp/SSjsuFVGmW3maK2af3ItJ94xnizKOZ+Pua79+7yzpLcoCOLzwW238e1/17KMUQz7y43MYQK92O7aXpmRMZZKUmbqRyns2ip8mU65tNRQ+FbCl0TnRvjSj+L3Q9u2yc8FcIIUCdY0xqa3kSeegDPPhFCIadPE9V1vLnLGQDbiq66C4cN5992o5JYRqNkyJdzyxsh2RVJET5pEVtFehrDe1OYcvYR7Dt0No0fDn/8sXhu/+MK0r8ZsF7dCXpP09MSCODyFnwLs2wfXXx+9vC9bKekkZtjGUvi+9AA76UlgZ0HM46k3OB4zTSyFf/HFyeUKtxJ+xrFCavBxhHaO2wQCxoxbfdly7r5bJB2zC3PsuupTrmUa3154J0NYz6ecj65H56zR+/RlAnNYdv0LjGEJ07APErc6bWMNlMGDje/W0FB5Tbt2NZYlYjJTi0qXlwvicjKTVFdHRbGa9iMVftu2zuslC/mgU+3hoDi+166HBx6AL77gie99ESmH+Oij5vWHIxy2jBjB1KlGKUMr7AhfRSgEM2ca/1sfRNKOP5HZJoV/ybqH6VKzF55/Xhw8Pd28IxpXAfhYkP0n0fkFHuGnAPfcA6+8Er38eL7jaK6wzcdS+H6/cO4Gdm2LeTxr+uVYiKXwk4Vq0gHILDnEYTqgu9zytDQopCOhnr0ILl7G00+LrJ8q4c+bJz67LXyHQjqwbOpjVCBGtl2HTUsTefc3jruRF7mRk/maANEXprpaPHBi2Z/toBL+xo2G/V/NGZOIwpfH9vnEvlu1srTr228j5+Cm8CXh+/3C55FqwpcOczvC1wjR4Vc3CZbOyaHjdOdaiCNYRSg9A044wfV4bnVwQbxIqERubRe9elF03EAmMSvyMBjARiatf5Y3sn4iAv5btRI7mjHDtGkswn/zTfjkE/d16gsq4av/NwY0e8K3SxTVly10Zw/7eooZJbEUvt8P28gjbc+2qN+sMePqzY2HZOS2dWXDlwMro7TQ1ZwDxrWqOXEU/lXCcRsMmk0648aJxnZc9CHvcwll1cbTUiYKu/NOY315bSsrYRmjyaSSUenRYUvV1WKASGJNZJCohD9okBFBlCzhy2NLws/ONtqlFx1Bz8/nAR6O7LekxDlaSPoApML/859Tl0lVKnw7k851/IPs5QtFJZrLLuMy3iELe+/ycFZTNSDfvTIKzjl4brlFzCN54gnzOlGED+wYOIlxzCdLqwB0/sxtVKa14pGsRzl8WIiLb3pOFnYnJb/3q6+K35x8O1ddJWY/NwbIMS0JP16V7yn8FMDOaTIBIQGvf1u8YsYi/EAgHL55YE9UknCfDyZMMP5PVOHLTmGn8Guj+iVpScWVGQfhRzrosJH4tnxHW45QU2Nj0pk1i0BZCe9wmelyqIrWus+jR2FxOCPnicGlUceWhC8JI1mFr0IlH7t7sXChcHhaSyDKY/v90Qo/fcNqtKoqrudl/AQjUTpR5gvM1yMrS2TR/H//T0w8TRSzZ4u6sirhOZl0Wpft5ynu5uiJZ8K118LVV9OGEi6yJKzNzwfQGcEqqge5R+i4ISMD+vc3v6GB/TVZkTuJbMrJ3fgll/A+E5nDpyc/zN6azpGaBU+tDafOVDJs3nOP+Ny1K+lm1husCj/evlwfbwItkvAnMpsd9ORbxCtsPCadjQwU/2zaFPW76lS1U/hpaSL/vqzzqsJpxioYaXaTgexk8mGWWVrIIZwnXaltqcwXjtuTWGFP+O+8Q7BtBz7jbBPhy6gUO8I/fFi8JRVquYzUowm/qsqs8FNB+LEU/iOPCNUtc8VIqBPWQiEL4W8SM726sZfz+JSqKsPOb4XqtFXbkozz9v774dAhs8NV9hnVjAIw9NU7aUUp3/7iRXHzzjyT7fTimkjFUYGZM6Ebe+hIIdVD3CN0JN5/XxT4UaEKpliE/8GRcVRraaR98j+e5Q42ZeSz7dybKS4WafMBNviGQPfuJrOOamZr7LASfmMwlRb7AAAgAElEQVRKG9EELl/tYCV8HzWczWfMZiKEC4HEY9LZwCDxT4xKGipRSYUeCIhwwwULoteXHdgphC5ZyE4nz79676G4FX75YOG4HcnyiEkigspK+PBDis+6mCBpJsI/fDia8OW1FSpaY23GaEbrS6KObTXpJEL4TnHwKskm8rYkjy3PrVUr4xpkfruGmnYd2ENXrudlXn7ZmfBVp616Teyc4LFgRx7yu0mwzJ5N17lv8Bj3ceS4sEjx+XidHzGR2XRlT2RVn88oWh4cGh/hX3wxjB1rXhYP4cuU1YercljX9jS0vzxPb3bQ+T/Pc+JocaPkw2zxEo1tA6fA7Nn4MQ+MZK5dfcMaPRVvX/ZMOimAVf2MZDkdKGIOhh0mHoW/mf7ofr85HtAG6s2VRNStm7HMelNlB463wEi8kO2QhJ9L/Db8qta51PTozSiWmdoIiNfso0c5NkUEzFvfTHTdPOjld6lqN7YZwxDWox8zn1xJifAvSmINhRC1Am+6KeaFcBpQajvs3qCcBpg1ckpV+FnfrqZy4Ahe5VrO5xNWT99NRUViCr82RUvsCF/+lkk53HwzFT378xj3RR5yFRUwjavxE+JK3jS1Q0boBAe7zA60wHquToSvmpr69TOSpK3InSQWXnkl7S86MzJGVfEw2zcZios5xWcWB7HMHpWVwvT17rtxnkwd4OBBkXFdXifPhl+PsCp8ab+fixHrGGsApqVBFRmUdzs+psK365DqNHUrObkRvpNyjQeqws+ijCwq4rfhB6Fq2ChGsty0rz59gHfegfbtqTpDXD87IlUHvby2ctLRdx1G4yeEvsKcoK24WDg2TQr/tdfE7KcYExHiIXw3hW9VjXJ/v/ud+OzYUezLRw1ZW9ZR3n8Yr/AT/IS4llfrReFHzG3KecgMBHLfv+L3sGULux94gUoyI+u+8w5s5gQWcXLYrCOYRSr8LfRFa9sm7rZYRVS8Jh1J+PN7XCVeFZ5+GjD6nerMXt91Avh8TNbN0TrWcaJpcNttxv/79wvT1+23x306KceBAyIhXKJvq9brWhdo9oSvdsbBg4X9fiUjOEjnuPchO25Jj0GOhL90qejDdvHebpNT6krhq4SfizHp6oUXnLeRpHLsGHy8dyTHs4V2FEXs34vmVYqZYBdfTFq2GKXWQtdgT/jy/La0Hw2Avtis3GwJX8ZXWo3sFtSW8K2w3qO8PLGvfmzBX1lOcZ/hbKUfczmbW7NeoaIsRFZWdPhvTY1h4qqtwpekqPaJN8NiPRCAwawXhUeuvprqM0QgvTxn6Qt6javJZ11E1fv9QuGvZnhCbYpX4VvFVjAo/opa9xLOgPBECXlualhlSaA9nHxyFOHb+WKef974bnpDbCAcOCDeMiJRbzEIX6ZeqY/i7s2e8NUb3y6tlNP40mTOiQeygx/pNlhUprZh9SuuEFxoV15PJXNrR3Sz4cvBHcvkZAfVaSsJ/xAdo8rWqZDH+cMf4KWVhuM2GBTXoMua2cI2c9llkXWVyLkI7NSsPL+S7C5sp5d4QiooLhYRM3JbX6Uy23JJtM1fhdPgjteGbx2Q1v979DDIEeD/fidG6MtcT9fyAkaXfEZWFlxwQXS77Ew6tVH4diIgEIBHuZ8SWsPTT0fl/XnjDfH5H35AFWkR562vvJT+bGYVIxIifDeF7xaWWV0t+oE1+tOuf4dCwOTJjGQZuRwy7UPCzgQir61dn3jvPXjuuejlqcaxY8I8KftyLJNOZiZMnhzbl5gKNHvCVy/2ydVfkE512GEbPyThF3YeJHb43XdR68jOFwmHpJxljGQyM/jqK2O9REw6cnAn86qnKvyOGJky3UKt5W9Hj8IKTgKEzyMYDA/Kd94RrHzOOZFB+umn0fuxU/jyvAMBWMpotGVmEj9yRCj8iHN0+ZdQWcm+rDz0JUtcJVttFb71+W2XIsDvh2GsIYifbxDTfN/nEo4GOnBl+ctkZkYrd6nwU2HSkWRgN3krEBAPo085Dzp1iiJ8+bwsogMfcQFX8iYBqglsWIsPndUMT6hN1hxRTgq/jcVKJAnfSvB2hF9TA0yZgg+dicyOLFfHiU3AXNQ4VHHppXDHHdHLUw3p04lH4R88KO6PLJpT12hRhH9OaDYVZPAFZyS0D0n4BzqG5/O7OG7l8UawipGsYCrvmX5PxKTjNLEmHjiZdOIh/OpqOEwuBeQximUEg5Dtr4T//U/YrdLTXd863Ew6kvB9BVtNlcXLyszO0eL35lBNgEfL70ArLhYJ3RzgRvgZVHAx7/NazZWEHvyt6XdJDlbCt+5PxpcPZzU7swZQiXgC6+mZzOj8Iy4KvU+ufiiKCFWFn+YzdpqMSUemNXjmmWi/SYAgPdjFdnoD0Zk9VR/SNK6hCweYxCwC60SETqIK39qHnAi/bVvzek4K365PhkLAyJEcIpcpGGYdVeEPGRK9nexn8to3RI3p8nIh0uJR+B99JD6326eYSjlaFOGPPDybLzktkgoAINfdjwkYhLu37cDwVMBoO7619J10eI7GbLpwyuxoZ5uUAzuZVz01SidewrdWp1rGqIjCP7vGMOeA+6TMWAp/SXgClpp3v7JStFVuO4E5LOIUPiOc2MXFjm/NYJlOJd/jI17XfsgBOvM+U/kB/0F78nFbT7hV/VvvUSAgzmMYa9jSyohmyc6Gf7e6ngyqGLPpX7aE3ypYzA3/ncj3nz0Z6SxNRuHLe1JQIGadqsgo3EOAGnbQS/xvIXz1DXEGUzhIR67mNXzrVlNEO3bQKyHCr63Cj8ekU1MjdjaLSUxmJhqhyD4k7FS87AuhkCjLGWPycJ0gEYVfH5E5KloM4XdhH10OrCX38olceKFYdu21wqMfC9IW+bO7skX6VpdIHUn4MqQxn7UiXC4M9ebX1BgvC3YqwJojPhFI87dK+Ifp4LqvSFhmmDiXM5J+bCWjrIiLqsLmnPC04ngVvvUNRiZo0zUtYmvQdTFIpFmkA4WcxApmM5ENDKLUl+Nqx5fX9M9/hvt4lP104SMupOvKT3mb7zORWXyPj9EqK0XqXQtiKXyfD9LLi8ljO5uzjRmp2dmwrGIoiziZ4Uv/jk8zj960Q3uZVTmOE7bPocuOZQxC3Gw3h+K0afYT7tSHkpVIM/cLeeik8E0zjknnLa7gIj7Av3AB63zDAa1OTDrxKnxHGz4wk8kcx36GISa8xbKHy3t3+HBUOp56g1T4MveQNVmdCo/wUwzZQc5B5HwdcddE+vcXy+xyfdjBZFIZPNjVpCMF5EiWU0o2AWoik1vAPNifeCK6nSpqQ/gPi1QvpKcLG34xbajxpbsqOTuFD5C3dxGTKz+Aiy6KjO54CR8E4aiEX0IbQscPiDhuq6vDOfszxQPiHD7Dh84cJhDCzzdZo1wVvnwTOlJQxMM8wEpO5Dw+4dv5+/gpLzOHiXzG2ejZ2SYWkG2KpfA1DXK2CsL5Ltus8IuLhfO2/d4NBJYuivx2PJs58/7T6Md3vPe9fwBwAeL9XT5gPv8c/vMf4zjr1wsRco1NQtHKSkOpWwkzY7+oxCYJPxAQbZbnZSXz17iaTCrRNnzDuoB4gCVCPPGadHJyzOslQ/izEDH70qwTK12J3Tiqb1KVYbqTJ4v/bTRGg6HFEP5EZlOe3QFGjIgMgFjZ/yRMJDlokPCwWGSgatLJppTBfMNbXAGYzToqmajT5FNN+BJS4ccy54DZhg+G4/acZY/TVi82qpOE2+SkCt0IP/JQOXGMUO26HiEmqUwnanMopg1LESGcqzLHivJ3DhnmbrsNfvtb6PntXPyEeICHmc55+LMMJqoig+rTz4bp0wFxvxaF+dka+WL3Ct5qiyB8q8I/elREvwSzckh79e8AnMRyvuQ0AuUlnMXnrDrxxxzocWIU4Z99tig6LiHfDmWKARWhkHP2xfR9QuFLk46mmWvoWs9nOSP5JjxzfH1ghO0+3RCvwvf5RLZViUWLEif8fXRlJSPiJny7e2cXOlxXUN9W09PhrLNE/p+Izti+3fQE8hR+iiGIRmcCc9g94Bzw+yMXOV7CB+X1dPBgMZIKzLnx5T7LyoRzz0+ID7mQfVpXE+E7VcSqD8KP9douB54kpCI6sJU+5Bd9wVGtbSSfuYTT24K1vbKuLhiDvWzIaDFLZteuqJwwE/TZfM5Z1CBWnl44RlyMVatwwkMPQdfVMzhCW77mZNv2VZ41Rczu2byZ0lLj+fGPf5jNOrY1X79bQyEdOJhmTJuWb4il5HDg7CvQ/vs2l/Ae8xhPGdnMfnAhyxgt0msPuYBTWEQuh6LeKCQhyftjRwK6bly7qIpVe7ZTkdORV/9jvLK6ET5okboEFz90IscfH20mcoObwrdeczUW/6abErPhy+swgymcxpfkUJIU4ae68IwbqqpEu6VVQEbfPPxw+J8+fRo0j3OLIPyBbKQHu9k9WBCWOmU+Xlx5pZhxGam8YTHrqApf2u+XMYq1WaMdFb7a8d0IvzaTSFTCj6UmrAofDLPOnFYXxe09tlP4ciDKULrnvhTqnaVLzYS/dSt9KDDNlVhMOHmLazy+zogDM5nDhMiDwko+5ePPFV+mTzdNktq0KTLpE7AnjazNq1nDMHSMp6ZqEiy69Hq0sjLe41K2kcepfMXhTgMi7dgy+AL8hDiPT6PudcdwTjvZ3pUrxdR8FaGQC+Hv3k7mgN6mORYq4VuPN28eVNzwc/jgAybcNYLNmxNzbsar8O3aKusemNpvQ/iVlYZ5dCaTSSPI2XzG3r3GOjIdslogyG4cpbIOQUGBqN7pBKt4icwr8SFMmLpusvF4Cj/FCAaJxPHuzxck4lQ4wg1paWEiHOSeRK20VNjvD/iOYy/d2NRmDAPZRFuOAGIAVFeL1/Z4FX4qCP8QHWPux2rDB6MC1ow2LjO2LHAz6cj6q0sqh4sDLl1qNumEZ9eqhL+H7uymmyvhD+YberKLGUxxbEd1z76iyMeMGVFT71XnfdR1CoXI3LyW1Qw3DVC1/1SPGA3jxjGXszmTBeylW+Stwe+Hfd1OYg9duYCP6NDBvHtJbOob2JEjUU0wEb7aDv/uHSKYQIEb4Y8bB398MZNI9EKCcFP4Vk1gJbR4wzJnzICrrxbfv+JUSsjhB/yH228XtQf+9CfjeqkPDLuHdW1SlFjRt697eutIpFj4Osg+6PcD69aJfxR/VH2HjbYYwv+OflT3EBWuklH4EcJv21ZkQ3NQ+G+8IRT+4nBt2A/3CSUrwzRrauDGG0VstGpGsHtVTQXhS6dtIgpfbddrXM1DPMii1pOi1nfanxvhP/ig+OzSO1PMKV+yxKyK5sxhl9aDTQww7WMxY9G/dnbcShvvTCabjqvi44+BKVPg888Z0lc89eXMy759jfWiHr5btuArL2MN5gRjKuEH0jT47DMmMJcjCHku75/PByF8fMz3mMxMzhhrX2/RzeRmJXyjT+j4dm13JfxUk4qbwreOqXgI3+m83wtPYakmnee5lSt5izOZz5o18POfG9dXvV9251qfpKoWz1E//X5gbbjwz/LlkcYnUnozFahzwtc0bYqmaZs0TftO07R76/p4KsrL4V//rGY885jNxKisfEkRPgizjkXhS9LKppSBbIwoY2kSkWadUEikEQFzR6wrhZ/pr6YtRykkN+Z+7Ew6++jKb3mIkD/6vTsZwm/VCkaOFPlGGD0ali2jslw0LDOtBubOZUHaBMDMAksYg7blu0i1EitpTGEG6xnMLnqajqvippuAc8+Figp+cvx8unUzVKR6/aMG4RrhsF3DMNM5q/Htfn/0AVUfTCgEH3EBbShh0AF7m4BbBJWcwCW/y/bmUohWVhZF+K1aGW8J6rlZSsUmBTfCtyIewo8HD/MAW+nDv7JvJB3xJLMjfLtxVJ/56NXiOVGf69YJZ0lFRYT8U13aNBbqlPA1TfMDfwHOBQYDV2iaNth9q9Th5ZdhLItpzTFmMzHiQKqNSUfXEWadDRtMvfngQfE5glX4CUWIvogObNH6mQhfEq+qbOqK8FtVCoKMR+HbmXQk7BzHyRC+zycIIhhE1DA9ehT9WzGLttPuVXD4MAvSo3MdqXZ8Gd53ww1iUTalnMkCZjBFZPR0aTPjxkFmJkN2TicQMJyKKsnbEb7u87GeIaYUQCpx2ZGYPGdJ+HM5h3IyGXPgI9uUFG5QnbY1Nca+exOeotmrl2n9kSOF5UDXxbojRwpfxaToF7WE4WbSAXG+UkzZqWu7a+VUOF2inGxu5m/0KtvEvTwOGPdJPca9NpIyGDT31bq0m1sJXz7Ec2qKReKpK0TknjTrNCvCB8YA3+m6vlXX9Srg38BFdXzMCNLShDmnBh+fc1aUwk8kR42pXNngwSJDkk29NemwlQofYGXAcNzK/Cpg7nh1RfhZZcYs21iwU/gSiUQKuRG+ponj1NQgFD6QtlLY5rt9I+z3CzOjCV+drCXNQjJH0Tjmk0EVM5gSk4TJyoKzziJ/1wwCAYOsXAl/9WqCfU8wzdC27t/uWCaTTkiQ1hwmMPbgx5x/fjTruBGRNSxTXs9eiBh8q8IfOlS8DBUViXXHj49ZozxuxFL4xcXhNzjs+5Kdk/YXv7A/lvrWM4vJzOhwBffzKAPYaKvwFy0iCsGgOaNmsuPJEphnCyeF37csbL+/4AKRSjNM+PUZMgp1T/jdATWf4q7wsgg0TbtB07RlmqYtOyhlcorQpo2Yor+MURyhfZTCTyRHjUn9ujhuR7KcPXRlH10jy9akj6YXO+nCPpPCVwe4nS0vJSadY8IbGau8IUSHZaqozdR7TTOXqPP7w4N00CBo1Yrs9eJh2GH5bMjP53Bal6h9ltCG8j6DYfHiSKibDLebwgzKyOILzjCRiTXSJYIpU+ha8i15oa2R+QRWf8rIkcr6a9YQHBxd81U9TzfC9/uNe/0RF9CtooDBRPcd6322ptVWbfhRCt9C+D3Dlq0dO8S1T2WKgVgKv1Ur4+35yitjbw/2pUjBiGCSeKzTs5SRzQvcxNdfiwsUy0Y/a5aoJSyRrIln//7Y6zgRfp/SMOHn54uSYWHCl3NAPjCXG64z1DXh27ljTDpG1/WXdF0fpev6qE6dOqXswE8/DWV7jjCWxZHsmLJTWUuQxQMT4cvQTBvCH8WyiDlHYn22ULKjWepI+HZP+pQQfmnjUPgq4QcC4UHn91PUbyT60qVkUk6rlQthwgTHY5UMEpO1aoLiwkmimcIM5jGeSjJNhO9IcueK8Mzx5dPRNLEfq8KXhNWao1BQQHBIdEWoeAnf5zOikz5GxBLKSVgqrMSl3ncnG35vtguGtYT+dA/Lqr17k7ebOyERG75anETCri1O+7Dmutod7MLdPMV45vNj/gnYE7h6DGtismSduPGMQ7mOvEbSbJtXslZkwOvZUxD+xo1w5AilpaIsQJIBUwmjrgl/FyheNOgBSlHNOkJhIdx9N6y8+w0C1PA+lwDGTai1wu/USUgPS6ROK46ZHLYSm1qdRA0+V8K3Cx2ThFGbKIO2NYkTvt0AclP4qt0cYtvwIyYd4JU1o+lVuJKz+BytqhImTnQk/OIBY+HQITqVinfr9HTow1ZOYDPzs6aYzsEV/fuzt1U/zigTkT2VlfD118bPVVWGDTof4VwLDY0mfPVYdm1WwzJvvBH+8hcYfE43NmSPtCV8p8R68jc7hT+m83Zhv7eEusi+XV4eXWe4tkiE8O0icBJR+FbCr6iAf3AdX3A6T3MXnThgOz5Uc631uiar8OMZh9Zi63KM9zq6TtjZNM0oCrx0aSRLbH2hrgl/KdBf07Q+mqalA5cDH9bxMcMDTedGXmQZI1kRJmB5E2pD+H//e3jGYP/o6lcnshIfOssYFbEZ3nknhLJasZ4hjGZppM4pxE/4iSp8dWZh4EjihG93PDvCKCwU9trOluJh1g5sJXxp0rniCpEqOZNKbuc59LQ0OOMMx/sy47DIstnv0JLIviYjwk6+yBaEn5YmEpDJ6z/CoTb38s7nMrb0s8jr3vz5xnmrCl8m7dLza2fSCQTgZz8TRDS/rZh12xGzCdON8K0zbeX1HJAVHZIJBoHKftVQCt8OiRC+tS9VVICOjxt4iRyO8Qx32hK4SvixCtzEi0QIX14jMcZ1epWsE+YciPiuWLyY0tJmRPi6rgeBW4GZwAbgbV3X17tvVXtUV8PJfM0w1vIiN0aWS7Xx1ltw3nnGa288kIR/332izOqKChmaabC2jLVfzkiuuUbc7KefFgNiKcJxG6rRIwNb7aipInxdN2cp9B0ppCY9kzJihyTFm/JYokMH4Sexqrjjjzf/b6fwS0vh3/82UiVPYjaccgrk5ESS21lx56tDKSOL3vuF/bO6WphzttKHHRn9I+dw9dVwyy1im2XL7AtaL8udQlaozEgripGev7raGITDWQ3t2qH17BG1j0Sctup6n7e6AB8652EO1bESirVSmkr4sgZP66LoSVdgEJ6swJZKhW9920t034kQvnXf0hy7kUE8zr38iNcZUzw7ajs7wu/Fdh7m1xzNP5UTOySegD6ecWhH+Mexj7bVhULhg7DvDRwIixdTVpZYtGBtUedx+Lquf6rr+gm6rvfTdf33dXmshQvF63l1NdzECxylNW9xRaRQgiTC008X6SwS6ajWyIIDuYOgqIjOHIgsG8UydtGd/Rxn6tSS8DtSSGDXtoiyVwe0nQ1fdu5ECN+6rr/oEMG2udi7U8xwI3w3k4787ayzxGeXLtG/WxW+TLGwjTwOhd8+tHCuHnUS1B/+YHwPksYKTmJwSXjGbVUV5zCXmUxG84nzs94nv99ega5sdxaVWkYkmRoQmbZfVSVMCbffDlcOXQPDhuEPRF+/WP4Cu1xIfj+sTz+R3XSLMus4KfyaGjFXRyX85ctFOGrG0UOuCv83vxGfcpJnKpBMPn8ViRC+9aGq5s97lPvZk9OfF3edxzzG8UseZzirAD2yPz9BRu3+gE84jwL6cD+P0nv3IsYVvZ9wu5NR+LfdBkMJX3xJ+AAnnywU/jG9+Sj8+sTGjSIz3+23Q6iwiO/zNm9wFaXk8MorMHdutK05EViJpKircNyq0RYjWR6x36sdNSODSObHRy42ArnVdewUvnwIJEL41tdbf1EhNe3iqPKCO6m7PRwlAfzud0LRWAlB3a9U+MrWkWsjc+2rJh0rESxmLIMqVhCgmmElX5JDKTOYEomQtXM4251XqZ7NqrbjTOmSZShhVZV4SDz7TIg229bC8OG253/66cZ3u9+dFH6wRovMupWTiMDZ1vzkk+Jz3jxjvaNHIU8Lh2RaYvAh+rqlMoGYvL/HHQf79iW+fSJOW6twUkm3kkz+MHkWT3E3bTjK49zHKk5kN915ePd1/J772UYe9y2+mOGs5hF+TR7b2EJfxjMv4XYn47S96SYYk21D+GPHwsGD5B4tSKqEabJoNoQvi0asXAk5771GFhURc07r1rEndsSClfB3txWEL4ta5FDCADZFInRUAkhPh7XkU0m6KZGaVPrXXVd3hO8rKiQUJ+G7IR6F7xRHrj4AogkfPuU8DrTKg1Hi2qmD30pcSxhDpl5BPms5o2wG1QSMqljE9j/89rfis6YGlnQ4FzZs4IdniNd7SdCS8CkoEPMthg2zJXQZnQvGOQ1TfLt2Cl86rD/iAlpzjHEYs26dbM3WeqehEJSUwMAs+5BMiL5uBw5ErVJrZGREv83FAzvCdxIUscxmZZ3z+BWPchIr6coeruWffMEZXFDzPvfyOGvJ55GR79Ob7fyG37GTXsxjPGeyAL3GeWC99ho8/rh5mZvC37ZN9HPp/FfHy5DQWoqzuohgD4mw43ZwyWLX2hKpRrMg/MOHxSsuQE1Qp93bL7KYMaxGeOxSYb+03pQdwW7orVtHFL502EqFr5JcRobIB7KKEYzBSAAmozhyc4Up6tgx8zFSQfhaUSGhDrUn/HgUvlM7rQrfuq/nuY3bL9gaGdEqWVmVn5xxO5bFjK+cyUJO5xitXduuHv+hh0Q7582DhTnC0fvwqdMj7ZdpedPTiaRUsFP4CxaY4/zlMZYuNdS0k0knGBSzbsvIMpl13Jy2KkIh8WLiNOkKoglfzQZaW7ilcY4HdsStLnvrLeO73w+XXOK83Zgxxvd9dGUa17L9if9w09SDdOQQ5zGdrzpfHMmgCjCP8XSgiNCqNY5tvOYa4a9Tx6TbOJwddiP87W9GuyUG1axjd4d88wb5+ZCVxZBji+u1DGOzIPw5c4x432FHF5JVsMHkrK0Lwg/WaDBocITw7WbYSkjSWspoRrIcH2IkS8IfOFB8fvuteTtpr0wkqiBK4RcegtzYk65iwe0axlL47iad8HK/8YR0I/zt9OYAnbiY/5Ffs9qULC1W+yT++1/x+faaAZCXR7tFBuGbsh2uXi3YbciQqH3062eu6CRJMD3dWO5k0qmpgQqymMOEMOHr6Lq701ZFUZHoKx3LtosddusWtY71Gp95pv2+kkGXLsLvmOxDxO7+d+woJketWCGKwoSte5HqXWA/Q9euOlh2NvgzAhQh5iZYiXo+4wDQP58Xs62SwMF9HMq2yTEbGS+hEAOC69nVdqh5g0AARo4kv9wj/IShktHUgy9STBv+ww8iyxKZJeoEa2errITiHoMjJp2RLGcnPThA9DtuRP0xmpxwcjW5DzCclDvDc5KXLhV1Omqr8DVCUHSYVr3q1qRTW4UP5mUq4UcPco3FjGUyswDocvUUXn/d+DUe1WmkQtbg3HNps3gWv+AZOHYsQvhpaQiF378/ZGfb+iWcnJeaJv7cFD4Is04e2xnMN6ZwXQkngpEVsXqzXaRdtbmgtXWsuiEjQzx0lAJoCcGOuDUN/vhHOPFE8zqxTDp256kWEIfo67qTXmyhL9r8eY5tlJFi6qVV92PtZ1KYyDEb2a6ggGy9jJ1WwgcYO5bBlSvJ0OovZWazIvwOFDKh+L/8ix9RRquo32sDa8TEQKAAACAASURBVCd97TX4/XuD6Mo+2nPYdoathBz4MgRR2vEluUinjfx/zBgR2lhbwm9LMVoohL+zIHyncEc3WPN62yERhS9z6Vhh9XlI2Ple5HXkuOO449XhXHWV/X4krO0yJaz65S8pHXYKz3AXk27oTeDR39GOIsOkMzw6/l6ehxv8fneFDzAdMeP3XKZTXe3stLUeS0YTje2y3dZh29gRz3i0I/x4x7GV8O3elOYxHt+XCxwHlzTXqQ9d67wIu/ZGKfxweNT2NhaTDsDYsWTolfQ5utrhTFKPZkX41zCNDL3SZM5Rf68N7FTJNwjH7bOXL2EA39qac8DocJsYwFFaRwhfEo+V8CVkno1kCT+XcGB5bi6HD4vX5UQhyTeVCj8W4UuFf+aZRkoCFZHMmZMnRw7+05+a26LClfB792b7Pz7jFL6icMBptHryN+ygF2d+cKd4zRoWPcPWek5Ov8dS+Lvpwca0oZzLdCor458gJAsm9dbsJ101dsTz9iHHW6zJbXbIzjaPV/V+y/3NYzxaUZHhp3Fooxr1ddNNxvd+/Yzvum4Uo5fHivSPcBrk7a1skgSHHbf9DzvXeUg1mhHh69zAS6zMPIV1mJ+mdWHSAdgQLgQ9cuMbADEVvo6P5YyMUviS4OxCCt1mv9pBJYlzhhuE37692eYcLyLxzC4PTelUUweBinhMOuo6btcD4GtOprjXUCOZPUbaX7t77Ur44W2+5hQW3vMhuz9dzUdcwNC54cooSRK+qvCtpKU+lNf3PJcz+IKKgyWReywnYrrZjP0ESTuwu0kSfjxmt3hNOnawKnz1fsu5ONKOH4l1tcCO8OXEPBBRORUVYp7D7NnG5D47hb8zrQ9lPpvB17Mn+33H0feQR/gJwe8XKXIHsonnKm60/b22sOts2+lNGVmcsF6U5oml8EHY8YezmnQqzaX9sM9SKdV/vPk/1PXaVIWN1daUgwkgHoV/880ixYI6YUqFNSxz2bLodewUvtM5l9CGJS+vNdl7JEHEox6tNUnludXUQGm/YVzFm3z81Eb4858jidasiHUcVeGr187vF8eZHPY1t/7+uaRTjT73swjhy0O63fPu7EYLhZot4cvxpjptreNYVsSyIivLWeHLN8ad9KImr68j4dtNjrTij38Uc0+uv95YZp14xbp1fJeZH9nP2rUiDPvoUUDTWOo/mT4HPMJPCH4/3MiLFNGOt4muvVpXJp0QfjYykPTqMujVi0OIOFtVCYBZJSxlNBlUMYw1jjZ8FW6/2cFE+NWGwk8W8Sh8TRMpFpxgVfjWuHLr/uVDRl43uxjyU04x/y/vTzwKX6khbdpGTVlQ2as/3Hqr/Y1XtnnkETHhz4pYCl/XxWTLY8NPo4Qc0j+bESELef7yf7uHi1Phk6aAVCl8p5xLHTuaHf/qLHY17UjVKeNFfK3N67O89k5vmWDMnZHBFir8fsSg3bSJrdlDI/t55RX45z/h7bfF/0u1sXQ+sjlSya2u0SwIP+PoQS7lXV7j6qgiFVB3hA+GWUdNoG4tUi1J+IwzjBm3o1kaZdKxI3XZqd06nt2xoP4IPxashB8r377VpGOXNdtqnpJkEA/hW6GmHZZvH+ec476NPM6vfiU4w+53N6dtKCSWZ7VNZw4TyPliOqEac9pnN5OOUx78xohHHjFqB8cLOxu+dQzKB6H1gdinjzmXTkmJ8V0l/MpTxotwI1lrVoEcR27jzu0tz+9H5A8JBtnRZmjE1CMfEvLefq0bldzqA82C8Dst/ZR0qqOctRKptuGrN1o6bs0VM8yQnWbMGCOO/Dlu58v9/ZjHONrc8kMe417y5z8PH33EADZGptxLwi8ri8+ObyL8YKE4ebWXJwg35RwvrIQfK9++fGDKmqwA349+cTNBkno8TlsJ+WBRFX51tdiH9aFtRSJROnZOW0n42dkiWidj33ZydolXH7dSkxKRSVdNQOH/6lfmAiSJKnwnk45toXCiTTrFxcZ3dSiUj3G248ej8N3g8xF5kOzukB95y1D7GsCS0ChCaJGCKHWNZkH4h86/hhPYxIYw+Vo7RqoVvvq6uI5wfK30tNlAdhqRJEnjSt7kWe5gqW8sGjqBJV/xC/7AxA9vgwsvZCODKCObrfRh2p6J/IWfcX+PaTFVJ1jCMqsPCXVfC7aWmfxSqfBjEf5xx4lPI17eOc2xhBxAiSh8WbJOteFXVzu/zalIJErHqvBB3CdJ+DMQM367rBATwOQbjXr+1v30Zjt6586J5fhuQKgPyEQJ/09/gp/8BC691LyOvB5y3889Z79vacPPzzdHfVUd10s4nmwIX46jo0fFfbTrszEV/rp1EAhQmHtCFOHLGd1HalpzsNMQj/ATgd8PmzEKdh53nHnQpprw1Tjxj/keb3//v8bUQBvIB4TMijeXCdzLE/xQe5NxLIAtW8nWKujCPk5mET/kXzzCr/mKU2mtF3MVbzCNa9kyb0fMdpoJv7BW5hwwEoSlUuHPnRu9jnqPJOGpzrZYOdflQE+E8Lt2NW8jFX48hB+P01YtgCIhv1dVGYS/k14U9xhMt9WC8KWVRk6wkseyEn5TUPd2aO2eCQMwzjkQEPfp5Zej88ZLYWB3fewwY4ZZ4YdCiGK/ajGEMKTCnzZN+IvULJ3WNtohQvgDBxLITmflStG3ZBvlWx7A7p5jjYrzdYxmQ/gqKivNNrxUmHSs2S8lQvhZP/BS14O88454rbVG+EnVoGkQSPdxgC4s5mTe4If8lof4IW9w6+glnIsgAlmMww1RJp1aEr5URLWZuWklfLtp/uo6aWmi5Jt0bMllbpCDx2nmpRtUG36qFL4ab28XSy4JXz7IduefS+dNC2jFMY47TmwjCV9CPbfebEdrAvZ7K7p0gXHj4l/fKQ7/nXeIpD1Xx1GsfamEX1ODIPyiIja+Y7bjq/6T5csTLzbu9yNMOkOHRl4gHnrIHO4p+8e+XmOF01ZGDNQhmiXhHzpkJvxUFn+AaLUZy7bep49wXNnl/JbE4aRgs7IMs1GihN82BYSfijwf6kCUaQessN6jDz4wT92PpfDdTDoyT78TklH4yZp05HlKtSf/35l/Lv5gFWfxOWlp4rY5m3R0YcNvYoSv6yKdcjziwWqXB3NfVJOmSSxc6L5PW8IPP33+dvk807rWcEw7wnc7j0B5iQjWHzo00jenTzcqsamEf6BP2HFbD2adZkn4AAcPuv+eKFTbn5W4450U5ZRDBNwJv4Q2FJAXqa/qBlWZJEP4a9fCeqUmWSoI3+pcs0OsexSL8GWqYjkJTIWmwZQpsduXiA2/NqkVwFD48v89fU6nOqMVU5iB3y+cxtZIPXnMjhwim/ImR/iJIFYtXLt+aS21aYXfbzYL1dQAvXrZ5se3RkglqvAzt4brZOTnR/hBnekeDBqEf6T7ENGweojUqcc8bXWHWGSRikRS6enClrh3b/KEr5JWWprZpudEMtIcsYZhCSp8nXbBQwlPuhpqk9SvtrA61+wQ6x66xfmDyBN07JhzfVA382hdKXy7OHor4UcUvy+D3QPP4dzV0/FpOrm5WmQ+h9VG3ZRj8BOFet9iEf6DD7rvy+83j90ZM2DwYJFm4RLeN0KnSAHhfxcWZ0OHRlKkqFAVvj/dLzImOs1cTCGarcKvCziZX5IhfBn9IvfpRKzSNLWWfAawKTovgAWyE2VTRoZe2ShMOnYK//zzzao7FoHa5dSxwq1UnFvd0GRs+PEofLWso4QM99y3z6zwa2pgUdtz6UsBOXu+pUOH6Al8UYQfp8JftCiu1RoVJEmq9zRWSUl1bsbll5vrFYC4firh33mn+JT58dV4/NqadDI3rxONz8tj2rTo36urhR8ici6DBjnXeUwhPMJPAE6FH5IhfPk9VoSB7ANrGEaAGtiwwfUYUpmoidNqg1QQvjXeHeDjj43qUxD7HqqE/9VXibfBbvKWRF1F6dhlu5QZS8vLzQo/GDTCM9t8OZ2srOhnu9xPooQfztHVpCAnS6kRPbEUvrqsd29hEjvBCN6LUvgSdnl14lH4vlCQLGxK1QFpm9YKr7LPxxVXwKmnmn//9luRkgRimytTiWZB+KmIwknkONaZosmadNR9Op2DXG9tOCHcZ8+5m3VSTfipuLbyLcVK6ol0dJXwrWkV4oEM4bNDXdnw7QhfhoLK48prUlMD+zLz2JY5EKZPx+eL7ldSaPRiByXkREvYJNvaGCGrhqmmvGRy41tDYu36nMyPrxJ+PAr/7Bn3cJgOvMxPGGrxrwU2rDPZR61vmO8rNdRrOUQTQrMg/PpW+NZJGHJgvvii+UZaYUf4TrMIJSRZbqY/FWSwfNpa9uxxPoZsS4Twa5E4TW1fbeBE+InU8pSnIdMgJ4q77nJ23Caj8GNBVfh2Tlu5XJ2IFQrB1+3PhfnzyQyVRfkA5L3tzXa207tpMnmckGNFJcNkMmdar73TBL55jDfl1Ymt8HUGfvMuB+nEFbzFWoYxmwmcxyd0Zj/awQNiplcYbiZFj/ATRH0RvowfdnLa3nADXHyx8/bJKHypbGsIsJ4hDGMNS5c6OyFlR73z6nBMX5K9KRkV7YR4CD8Wd7VrJ8KU1ZJziaB1a/jXv+x/S8aGHwvqudo5bSFa4et6mPArKxly4HPHN8febGcHzdth+9e/whNPmBPTJUP4Mgzzgw/E9e7ZE958M3q9eYwXNqBwwRIr4cs3DokBbKJd8Q4e5gF6sIt7eYyBbOQTvscawhNuXBS+CjffU6rhEX4CeOkl0R+sDsRkTDqyzbFs+Opb+1ryyWctF18Mb7xhv75sy0l5tTPpLFgQ0z8cNyThW88xUf9Av361u9dOPrG6UvgSKuGr7bc6bXUd1rQ7E7KzGbFvhqNJJ6LwmzE6dYJ77nF/WMaDf/8bHngALrjAWGaXJmEe48WX8DRwq0lHTYEMMIUZAMxkMkV04AnupQ8FXMGb4t60bm3Ua8Sd8OshOCeCZkv4X36Z+uNkZAg/jLUzJEP41uyO6jmonVN9+q9hGN3YS0cOsnatKGdqnYAilUngSJjwY2UBc0AgkDpnkgwtrYscR4nA6XziJfwZM+I/lnV2sYS1ZJ/qtA2FIOjPgLPPJn/3dFvCf+6RY+RyuNkTvh2SCSDo2VPkrFcfHKoPTj5Ed9FTKPL//Q9wz1QKMJmZbGQAO5T7ECSNf3MFY1ks3hYUsWVH+P36iZIO8USgpQrNlvBdcpnVGuosXki9wrfOTJWQjtt81hIIiKn3S5cav+u6YbYIFBcKj1cq5Got4WTSUf+vhzQiZGTACy9EL4/XaZtIG51MOlaFb02m5fMB555Ll5ItHF+53rR9KAQ/+57Ip9QSCb/MPiAmYaiEb8qRc+ml8MUXVGzf73iv778fMqhgHPOZyWTTbyedJL9FF262I/zS0voXPc2W8OuS56y52JOZaSvbbGfDt/pZ5fGkbTCftTz6aPT+X39dlFsDCBxJfNKVHVLptLW+hqv3TU1hW5e40SaDdrwKPxHCd3qAW80SmmZOmaxpwEUXUZ7Whr8XXgK7d5uO698lQjKbuw3fDtZ5CclCJXxT2o2pU0HXqXz7A8dtQyE4gy/IpjyK8N3GihPh11eEoUStDqdp2mWapq3XNC2kadooy2/3aZr2naZpmzRNm+y0j1TA6SmZnW02j6QKyRK+7BDHH28MfKvCHzkyerbrT34iPg/QhQN0cpxxq6aT8B2pfR6dVEESvtV2qiY127u3/trz3Xeww5J4VIZBuhF+IsXk43XaynVNCr97d/503gw6h/bB2WfTtlSEZek6+HYKwm+JCl/WLa4t1BxNpvQ1+flw/PH4P3jXdrvHHxd9YDIzqSTdiN+PA81F4a8DpgIL1IWapg0GLgeGAFOAv2qaVmen5vSULC2FDz9M/fFkymCJWPY+FUeOwBdfRCt8+f+4cdFKUiWMeFMs+JsA4bdvb0y+SqbAerLo10/YdlWkmvCdbPh239UqWPJe7+h+Cj9oMwP27OGu6WdzHHuFP2f7dqoJsBcloN8BqSLIxoJYuXLiRffu8NprNj9oGpuHXUrGl5/RjqKon9u0CdcjZiZfcAZlxB9eY0f4SiaHekOtDqfr+gZd1zfZ/HQR8G9d1yt1XS8AvgNs8ts1Tdx6q/D+SyRCBG3bEkl/C9EmnbQ0d9PBGoYxhPX4cH/K+IoaH+HblTb8zW/EtXz88fptkxUy901d2/DB3mEv69yqb3yL/afC9Om0O7aLzzibxR/uhx072ElPLrw4tn769NP4ayE3FSxaJKLlagunlNlXvncpaQS5kGil6PdD66O7yWddxJxz++3w2GPmWeN2cIrSaVKE74LugFrad1d4WRQ0TbtB07RlmqYtO6jaJBKAvHk/+lFSmycMTTPntk/G4Wg16ch9ZGW5K/y15JNNOf3YErVPdbvGSPhO5eJ+8IP6Vfh28Pvjt+HHYyZ0suHLY6nrWOvcyt9CIeD005lUM52e7OSEG8+C1avpdVovU60At3NqBD77lOLkk6Mn3+3YAVuih4MrnAh/GaPYQU8u5V1bn9MJBTMBIw1GIAD33iuEixucCL/RmXQ0TZujado6m7+L3DazWWZLi7quv6Tr+ihd10d1ckt44gKpih97LKnNa41ETDoS1kEvZ/Ll5EQTvjpDVHXcOqEbu/GVHIVu3RJvWB0g0ULsDYF4TDpudXOtsEbjqIhX4Uu7/kLO4Hw+ge3bYf16Av16Nzsirw169kw8lt05Ll7jPaYyiVm0psT0i98PA7bNZDfdIjUq7Aj7mmviP16jU/i6rk/QdX2ozZ+zK1soetVK2gNwSQiQGrhNbkg1VFJOxKQjYQ3LlCFnrVtHE/6kSXD11eL7NwymBp+rHf+HvC6+TJ2aeMMssJsnkCisYayNEYnY8OMZpIko/CinLcYbx7Fj4v/Jvx8Hn3wiOrnVq+8hYbhVQXuXS8mkkvO1T03LA1oNA3bOZhaTkJrWblxceGH0stNOsw8Vb3SEnyQ+BC7XNC1D07Q+QH+gzrP71yfhq6iNSUfecJkOtnVr8wNk8mTz+hVksZn+LoSvczWvERx7qpGasRa46ir42c+wDQONF02B8KurRYWpeBR+bQnfeu/tnLbyAXTkiPi/UydEOb7du428vh6Shhvhf8Wp7Pv/7Z1/sBXlecc/z+XCvRGs/DIxCGhIEAUkiFcG46TFQnPRZmDoJR2gjtpmdATjpFU7NUPSdOo401Qdkk5FS0cnGJVEY1UaoyiJmTSmaHGwyA9RkhjFMAYpUlNF7pWnf+x5OXvP3XPOnrO/zp59PjN3zp7dPbvPe3fP9zz7PO/7vHyMpTq4t86pv97GyA8OD+qOGTQYLOhHYOxYuPvucPsmSdRumUtFZD9wIfC4iGwGUNVdwIPAbuBJ4FpVbSLw0Rhplhn1i3yt+jnVqObhjxpV7o2wdm25K73/fDuYFRjSUYXzeYEZ7GZg5eWNGxVAdzfccUfTA3ZPHKPVOXrUG7QWxsNPMqRTGcN34xNOjMYcPTp9t7ANqSX4xxnGIyxl0fEf0k25atqEnZtREZ7mj06s819nd99UGxHsv2wuZ5UrD19VH1HViarapaofU9Ve37ZbVPWTqjpNVZ+Ibmp9sigeOHWql3RslFohnSuv9GrlXHddeX9/nuAlzuVT/IKR/G7Ica9gA0fpQr/QhFEJkQfBd/T3V//CNhLDbzZpWxnDDyoTbEQnSPD9UxD+G3/CKP6PXjafWPfxHZuRnh4u/8tyZwi/4FcbUR60b7VyI0ljrkKTuIkZfPWRGsKJir9SI3g3QkcHrFw5+Gbw1+9xidsZ+CafBToGjrGCjTzGEjrGpligow55EPzVq71OTQMD1T18N5nG/Pn1j9dst8xKD991q0zz6bUIBAl+X195+SfM59ioMfThhXVGc5hxr26F3t5BCWK/c+Du82o5Pf894a5nrjz8VuShh+DFF5M/z+TJ8MwzcM89zX3eXfywk4L4Bd/V1KmM40/Z8zjjOcQGrkjdc6hFrcfnVmH48HJPqWqCP3u211Fm9er6x6s22AqGevgjRnjC7vfwXdLWXfc4Zh4zygTdk/7rNMBwRixbQt+If2c4x1jAj5Djx6G3l9WrYelSbz//98xVY61WZdb/VP/mm97yjvpjKGOl7QR/2TL49KfTOdf8+c3Xsq708irXV+Lv0vgaZ/Iuo4YI/owXNnCA03iKz7VUmDcPHn5nZ33BB++HvtEYfj0Pf8QITyQqu2VC2cO3bpjxEiT4Q65rXx8nHTvCH/JjetlM/8hTYN48hg0rzy7p/565+7ye4Puv5SuvNGV+07SQLBSLah5+tS+238NXOtjJzMGJ24MHmbLnce7nz/iQThP8BvGPcE6yHj4M9fC7ujxhrwzpQFnwzcOPl44OWLNm8Lohg7cWLuTo8JPp42F62cyh2QtOXAh/wT1HM4KfdqiuhWShWFR+6R3VxKZy0FK5pk5JpTZuZNjxATYQMOojY/Ii+EHLzRKml47bp6vLE4nKpC2UxcM8/Pip2/Osu5vdU/6Yy7iPybzBoZ5yd0wn1P7vpQvpDCq57MNfPqXyOGlhgp8RlV96R7Uv9tq1g9/vYBZjOcwEN57t3nt5a8J57OTcoR/OmDwIvt+DTtrDrwzpOMEP8vBN8JMjzFPTrml9fARPwd+ZWxZ8FxLyz3X71a/CggXVe+25+yBu56IR7EExIxoN6Zx99uD3gxK3u96BF17g0XO+mcJ45sbJg1gl6eFXCn5lD40gD9993gm+hXTiJ8x13jvlEt7jI7zOZPonlEtSOyfG781PmgRbtlQ/lutanaXgm4efEY2GdCoZJPgbNqCdnXxtz4oYLYyPLMZHNEqSHn7lNXYjwsN4+Ja0TY4w/9P+ESO5kdv4GjcH9rmvFr4Jwl33Sy4J/5m4McHPiEZDOn6efBLeYQxvMJHz2A733cfbF1zCQWIqGF5A4va6aoV0XDiglodfGdIxDz9+wlxnEbiT1XyfLwQOnPKHdOoxdiy8+iqsW1de10zhxSiY4GdEFA/fTcu2g1ks5RE4cIC3elsvWZsn0gzpuORekIdvSdv0aPR/6r+OzXj44M1250/UmuAXhGrDsMN4csOHe5X3XuJcujgGY8Zw6MLPx29kgUgzpFM5x6/fw6+sTmoefnKE+Z/6Rd7fI+ecc7zX88+PZoMJfkFw8bzKYdhhHzNvv71cYoEVK5DurngNTIAZM7K2oDpBRbDiOl6lh19N8G3gVbqE/a45/ILf0wMvvwxf/nJz595VqoqStuCb35ARboRu5SNh2ARnZ6dX72M7sznv2msZNnQKzpbi/ffTLxTVCL/z1aFLOoZfGdIZPtwTk1rdMs3Dj58ogg8wbVrz53ZVcM3DLwjOw682Kq8enZ1wgAnMYTtMnx6fYQnR3d3aXqqrOw/pe/huAhQbeJUujf5P45yxzZ3bP4I+DUzwM8IJfqNJH8dHKzrkpO0ptBuHfU9IScfwK4/f2RlcLRPMw0+SRv+nQTNWRT23hXQKQrWQTljOOAM++1mvhvfZZ5eTSEZzzJ1bXk7aw6+swzJsmCf2H344dODVsWPe/q1UG6ldqHWdn33We3XX4+ab4bTT4ju366kT549IGOw2yohKD/+22+DCCxs7xpw5nqewdy88+mi89hWNlSvLtVXSKp7mF3zwvPwgD9+8+2Sodp2vugo+8xlv2Q2SWrgw3nN3dcFzz6X/vTXBzwg3cMM9st9wA/z8540dw82KZERHBE4/3VuOI7lcK6QT5OGDJ/iVMfz3389HLaI8Uk3wv/Wt8vJFF3lPX/PmxX/+uXPhlFPiP24tTPAzwvXUiCLYHR3NJ32NoTiRbWZS+kpqhXTc+0rB7+8vb3M5mj17TPCTotqTUx4m7GkWE/yMcDG8KFn6jo54ew4UnaAa583iF5N6MXy3b39/eZvLKbz6qgl+UhQxL1LAJrcGzsOPKvhGfMQp+P7h843E8N2+/knLTfCNuDDJyAgnCFHExQQ/Xtat85J1M2dGP5Y/Phyml45777Z1dpaPYYJvxIXl/zOimZlufvazwe+DBP/227Mtv5pn5s4td8eLSq2ePtUEHwb/OJx0Ehw5YoKfFHHkavKG+YgZ0dVE6ZuLLvL+HEGC39dnffJbgTCC7xL21aZDdMlDE/xkcL2yVq0qr7v11mxsSQvz8DMijrksgxK21me7NWjEw6+W4HVjNUzwk2H8eK/b68AA3Hmnt+7GG7O1KWlMHjIiDsF/4IGh60zwW4Nagl8taQuDBd88/OTp7i6HdooQCrWQTkY0E9KpJOjxs5UrUhaJWj/otWL4/pCO8/Ddq5EMIvDmm/Dww1lbkjyRBF9EbhWRl0Vkh4g8IiKjfdu+IiL7RGSviPTWOk4RicMT7w34r5qH3xrEkbR1Hr6LNRvJMWFCew+4ckT18J8GZqrqLOAV4CsAIjIdWA7MABYB60TEfE8fcUzsPWrU0HUm+K1BHEnbd9/1XidOjNc2o7hEEnxVfUpV3dChrYC7NZcA31XVD1T1V8A+YG7QMYqOf4BNowSFb0zwW4M4krauRr+bLMMwohJnDP8vgCdKy6cDb/i27S+tG4KIXC0i20Rk28GDB2M0p/XZsgVeeineY5rgtwbNJm39Hr5LJsaR4DcMCNFLR0S2AEGVoNeo6mOlfdYAA8D97mMB+wcOc1DV9cB6gJ6enkINhViwIP5jWtK2NYgjhh/mWIbRCHUFX1VrVoIWkSuAzwMLVE+MXdsPTPLtNhH4TbNGGuGJIzdgRKeWSJ91lvc6e7b3Ws/Dt6c2Iy6i9tJZBPwNsFhV3/Nt2gQsF5EuEfkEMBV4Psq5jGCmTMnaAiOIWmGYiy/2QnnXXOO9Nw/fSIuoMfx/Bk4GnhaRF0XkLgBV3QU8COwGngSuVVWbqiMBnn8exo3L2gqjkso5hyuZObMs7n4P/q23hu5rHr4RF1F76XxKVSep6uzS3zW+bbeo6idV6pp4wgAABn9JREFUdZqqPlHrOEbzjBsXT3VHI14a6Tvv9/D9g39cSMc8fCMubKStYSRAIyLtF/y33x663Tx8Iy5M8NsAS9TmG7/g339/edk8fCNuTPDbABP8fOMX/GXLhm43wTfiwh4W2wAT/NZk40bYtq3+fv6Qjb93j3XLNOLGbiXDSIjly72/etQbLGcevhEXFtJpA8zDzzf1BN88fCMuTPDbABP8fFNN8C1pa8SNCb5hZEw9Dz9o7mLDaAa7ldoA8/DzTbWQjfPw7foacWGC3waYIOSbeh6+XV8jLkzwDSNjqgn+TTd5r2PHpmeL0d6Y4LcB5gHmm2qCf911XlinCHOtGulggt8GOMH/xjeytcNoDpu0xkgLE/w2wAm+Vc3MJ9bP3kgLE/w2Qgs1QWT7YB6+kRYm+G3A9dd7rxdckK0dRnOY4BtpYQ+TbcDChebd5xkTfCMtzMM3jIwxwTfSwgTfMDLGkrZGWpjgG0bGWK0cIy3sVjOMjLGBc0ZamOAbhmEUBBN8wzCMgmCCbxiGURBM8A3DMAqCdQgzjBZg3Tro6cnaCqPdMcE3jBZg1aqsLTCKQKSQjojcLCI7RORFEXlKRCaU1ouI/JOI7CttnxOPuYZhGEazRI3h36qqs1R1NvAD4G9L6y8Bppb+rgbujHgewzAMIyKRBF9V/9f3diTgSngtAe5Vj63AaBH5eJRzGYZhGNGIHMMXkVuAy4EjwMWl1acDb/h2219adyDg81fjPQUwefLkqOYYhmEYVajr4YvIFhHZGfC3BEBV16jqJOB+4EvuYwGHCizgq6rrVbVHVXtOPfXUZtthGIZh1KGuh6+qC0Me6wHgceDreB79JN+2icBvGrbOMAzDiI2ovXSm+t4uBl4uLW8CLi/11pkHHFHVIeEcwzAMIz2ixvD/QUSmAceBXwPXlNb/ELgU2Ae8B/x5xPMYhmEYERFtobnxROQg3g9HM4wH3o7RnDxgbS4G1uZiEKXNZ6hq3SRoSwl+FERkm6oWanC6tbkYWJuLQRpttuJphmEYBcEE3zAMoyC0k+Cvz9qADLA2FwNrczFIvM1tE8M3DMMwatNOHr5hGIZRg9wJvogsEpG9pdLLNwVs7xKR75W2PyciZ6ZvZbyEaPP1IrK7VIr6RyJyRhZ2xkm9Nvv2WyYiKiK579ERps0i8qela71LRB5I28a4CXFvTxaRZ0Rke+n+vjQLO+NCRO4Rkd+KyM4q25MtLa+qufkDhgG/AKYAI4D/BqZX7LMauKu0vBz4XtZ2p9Dmi4GTSsuritDm0n4nAz8FtgI9WdudwnWeCmwHxpTefzRru1No83pgVWl5OvBa1nZHbPPvA3OAnVW2Xwo8gVePbB7wXJznz5uHPxfYp6q/VNVjwHfxSjH7WQJsKC1/H1ggIkHF3PJC3Tar6jOq+l7p7Va82kV5Jsx1BrgZ+EfgaJrGJUSYNl8F3KGqhwFU9bcp2xg3YdqswO+Vlk8h5zW5VPWnwP/U2CXR0vJ5E/xqZZcD91HVAbyyzeNSsS4ZwrTZzxfxPIQ8U7fNInIeMElVf5CmYQkS5jqfBZwlIs+KyFYRWZSadckQps1/B1wmIvvxSrZcl45pmdHo970h8janbZiyy6FLM+eE0O0RkcuAHuAPErUoeWq2WUQ6gLXAlWkZlAJhrnMnXlhnPt5T3H+IyExVfSdh25IiTJtXAN9W1dtF5ELgO6U2H0/evExIVL/y5uGHKbt8Yh8R6cR7DKz1CNXqhCo1LSILgTXAYlX9ICXbkqJem08GZgI/EZHX8GKdm3KeuA17bz+mqv2q+itgL94PQF4J0+YvAg8CqOp/At14NWfalURLy+dN8P8LmCoinxCREXhJ2U0V+2wCrigtLwN+rKVsSE6p2+ZSeONf8MQ+73FdqNNmVT2iquNV9UxVPRMvb7FYVbdlY24shLm3H6U0q5yIjMcL8fwyVSvjJUybXwcWAIjIOXiCfzBVK9Ml0dLyuQrpqOqAiHwJ2IyX4b9HVXeJyN8D21R1E3A33mPfPjzPfnl2FkcnZJtvBUYBD5Xy06+r6uLMjI5IyDa3FSHbvBn4nIjsBj4E/lpVD2VndTRCtvkG4F9F5K/wQhtX5tmBE5GNeCG58aW8xNeB4QCqehcJl5a3kbaGYRgFIW8hHcMwDKNJTPANwzAKggm+YRhGQTDBNwzDKAgm+IZhGAXBBN8wDKMgmOAbhmEUBBN8wzCMgvD/Wqth2QrXe38AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f016d16ccf8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Downsampling in time through FFT\n",
    "\n",
    "t_sample = 50\n",
    "X_train_sub = signal.resample(X_train,t_sample,axis=1)\n",
    "if (N_val > 0): X_val_sub = signal.resample(X_val,t_sample,axis=1)\n",
    "X_test_sub = signal.resample(X_test,t_sample,axis=1)\n",
    "\n",
    "# show a downsampling example\n",
    "plt.plot(np.linspace(0,1,T_train),X_train[5,:,5],'-b')\n",
    "plt.plot(np.linspace(0,1,t_sample),X_train_sub[5,:,5],'-r')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2458, 50, 22)\n"
     ]
    }
   ],
   "source": [
    "print(X_train_sub.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# # normalize data across the time sequence dimension only\n",
    "# scaler = preprocessing.MinMaxScaler()\n",
    "# for i in range(N_train):\n",
    "#     scaler.fit(X_train[i,:,:])\n",
    "#     X_train[i,:,:] = scaler.transform(X_train[i,:,:])\n",
    "    \n",
    "# for i in range(N_val):\n",
    "#     scaler.fit(X_val[i,:,:])\n",
    "#     X_val[i,:,:] = scaler.transform(X_val[i,:,:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# def downsample(data, mul=10, enableTrunc = False):\n",
    "#     assert data.shape[0] % mul == 0 or  enableTrunc, \"Can't downsample without data truncation.\\\n",
    "#                                                 \\n Set another value of mul or set enableTrunc = True\"\n",
    "#     outdata = []\n",
    "#     for i in range(data.shape[0]//mul):\n",
    "#         outdata.append(np.mean(data[i*mul:(i+1)*mul,...], axis=0, keepdims=True))\n",
    "#     outdata = reduce(lambda a,b: np.append(a,b,0), outdata)\n",
    "    \n",
    "#     return outdata\n",
    "# span = 2\n",
    "# X_train_sub = np.expand_dims(downsample(X_train[0,:,:], mul=span),axis=0)\n",
    "# for n in np.arange(1,X_train.shape[0]):\n",
    "#     X_train_sub = np.append(X_train_sub,\n",
    "#                             np.expand_dims(downsample(X_train[n,:,:], mul=span),axis=0),axis=0)\n",
    "\n",
    "# X_val_sub = np.expand_dims(downsample(X_val[0,:,:], mul=span),axis=0)\n",
    "# for n in np.arange(1,X_val.shape[0]):\n",
    "#     X_val_sub = np.append(X_val_sub,\n",
    "#                             np.expand_dims(downsample(X_val[n,:,:], mul=span),axis=0),axis=0)\n",
    "\n",
    "# print('X_train_sub: ', X_train_sub.shape)\n",
    "# print('X_val_sub: ', X_val_sub.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Expected input shape: ([batch_size], timesteps, data_dim)\n",
    "# Note that we have to provide the full batch_input_shape if the network is stateful.\n",
    "# the sample of index i in batch k is the follow-up for the sample i in batch k-1.\n",
    "\n",
    "# perhaps should try masking layer\n",
    "\n",
    "data_dim = C_train\n",
    "seq_split = 1 # Set to one when using FFT to down sample\n",
    "seq_len = int(X_train_sub.shape[1]*seq_split)\n",
    "timesteps = seq_len\n",
    "num_classes = 4\n",
    "batch_size = 200\n",
    "num_epoch = 100\n",
    "\n",
    "model = Sequential()\n",
    "#1\n",
    "model.add(GRU(200, return_sequences=True, stateful=False,\n",
    "          recurrent_dropout=0.6, dropout = 0.6, input_shape=(timesteps, data_dim)))\n",
    "\n",
    "#2\n",
    "model.add(GRU(100, return_sequences=True, stateful=False,\n",
    "          recurrent_dropout=0.5, dropout = 0.5))\n",
    "\n",
    "#3\n",
    "model.add(GRU(50, return_sequences=True, stateful=False,\n",
    "          recurrent_dropout=0.4, dropout = 0.4))\n",
    "model.add(Flatten())\n",
    "\n",
    "#4\n",
    "model.add(Dense(100))\n",
    "model.add(BatchNormalization(axis=-1))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dropout(0.5))\n",
    "\n",
    "#5\n",
    "model.add(Dense(num_classes, activation='softmax'))\n",
    "\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='rmsprop',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "# define early stopping callback\n",
    "earlystop = EarlyStopping(monitor='val_loss', min_delta=0.001, patience=30, mode='auto')\n",
    "\n",
    "# saves the model weights after each epoch if the validation loss decreased\n",
    "checkpointer = ModelCheckpoint(filepath='GRU_model.hdf5', monitor='val_acc', verbose=1, save_best_only=True)\n",
    "\n",
    "callbacks_list = [earlystop, checkpointer]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 2089 samples, validate on 369 samples\n",
      "Epoch 1/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.8997 - acc: 0.2413 - val_loss: 1.4345 - val_acc: 0.2846\n",
      "\n",
      "Epoch 00001: val_acc improved from -inf to 0.28455, saving model to GRU_model.hdf5\n",
      "Epoch 2/100\n",
      "2089/2089 [==============================] - 4s 2ms/step - loss: 1.7125 - acc: 0.2829 - val_loss: 1.3570 - val_acc: 0.3631\n",
      "\n",
      "Epoch 00002: val_acc improved from 0.28455 to 0.36314, saving model to GRU_model.hdf5\n",
      "Epoch 3/100\n",
      "2089/2089 [==============================] - 4s 2ms/step - loss: 1.7017 - acc: 0.2652 - val_loss: 1.3413 - val_acc: 0.3848\n",
      "\n",
      "Epoch 00003: val_acc improved from 0.36314 to 0.38482, saving model to GRU_model.hdf5\n",
      "Epoch 4/100\n",
      "2089/2089 [==============================] - 5s 3ms/step - loss: 1.5722 - acc: 0.3112 - val_loss: 1.4055 - val_acc: 0.3577\n",
      "\n",
      "Epoch 00004: val_acc did not improve\n",
      "Epoch 5/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.5806 - acc: 0.2954 - val_loss: 1.3940 - val_acc: 0.3767\n",
      "\n",
      "Epoch 00005: val_acc did not improve\n",
      "Epoch 6/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.5546 - acc: 0.2978 - val_loss: 1.3961 - val_acc: 0.3577\n",
      "\n",
      "Epoch 00006: val_acc did not improve\n",
      "Epoch 7/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.5284 - acc: 0.3107 - val_loss: 1.3503 - val_acc: 0.3957\n",
      "\n",
      "Epoch 00007: val_acc improved from 0.38482 to 0.39566, saving model to GRU_model.hdf5\n",
      "Epoch 8/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.5001 - acc: 0.3131 - val_loss: 1.3741 - val_acc: 0.3631\n",
      "\n",
      "Epoch 00008: val_acc did not improve\n",
      "Epoch 9/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.4960 - acc: 0.3241 - val_loss: 1.3471 - val_acc: 0.3821\n",
      "\n",
      "Epoch 00009: val_acc did not improve\n",
      "Epoch 10/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.4695 - acc: 0.3207 - val_loss: 1.4275 - val_acc: 0.3604\n",
      "\n",
      "Epoch 00010: val_acc did not improve\n",
      "Epoch 11/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.4238 - acc: 0.3375 - val_loss: 1.3877 - val_acc: 0.3550\n",
      "\n",
      "Epoch 00011: val_acc did not improve\n",
      "Epoch 12/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.4211 - acc: 0.3447 - val_loss: 1.3216 - val_acc: 0.4282\n",
      "\n",
      "Epoch 00012: val_acc improved from 0.39566 to 0.42818, saving model to GRU_model.hdf5\n",
      "Epoch 13/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.4322 - acc: 0.3289 - val_loss: 1.3428 - val_acc: 0.4038\n",
      "\n",
      "Epoch 00013: val_acc did not improve\n",
      "Epoch 14/100\n",
      "2089/2089 [==============================] - 7s 4ms/step - loss: 1.4043 - acc: 0.3394 - val_loss: 1.3713 - val_acc: 0.3821\n",
      "\n",
      "Epoch 00014: val_acc did not improve\n",
      "Epoch 15/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.3835 - acc: 0.3576 - val_loss: 1.3965 - val_acc: 0.3821\n",
      "\n",
      "Epoch 00015: val_acc did not improve\n",
      "Epoch 16/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.3694 - acc: 0.3643 - val_loss: 1.3850 - val_acc: 0.4146\n",
      "\n",
      "Epoch 00016: val_acc did not improve\n",
      "Epoch 17/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.3621 - acc: 0.3571 - val_loss: 1.3617 - val_acc: 0.4228\n",
      "\n",
      "Epoch 00017: val_acc did not improve\n",
      "Epoch 18/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.3819 - acc: 0.3475 - val_loss: 1.3668 - val_acc: 0.4173\n",
      "\n",
      "Epoch 00018: val_acc did not improve\n",
      "Epoch 19/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.3212 - acc: 0.3815 - val_loss: 1.3591 - val_acc: 0.4282\n",
      "\n",
      "Epoch 00019: val_acc improved from 0.42818 to 0.42818, saving model to GRU_model.hdf5\n",
      "Epoch 20/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.3170 - acc: 0.4059 - val_loss: 1.3584 - val_acc: 0.4173\n",
      "\n",
      "Epoch 00020: val_acc did not improve\n",
      "Epoch 21/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.3326 - acc: 0.3667 - val_loss: 1.3436 - val_acc: 0.4472\n",
      "\n",
      "Epoch 00021: val_acc improved from 0.42818 to 0.44715, saving model to GRU_model.hdf5\n",
      "Epoch 22/100\n",
      "2089/2089 [==============================] - 8s 4ms/step - loss: 1.3215 - acc: 0.3839 - val_loss: 1.3132 - val_acc: 0.4499\n",
      "\n",
      "Epoch 00022: val_acc improved from 0.44715 to 0.44986, saving model to GRU_model.hdf5\n",
      "Epoch 23/100\n",
      "2089/2089 [==============================] - 8s 4ms/step - loss: 1.3067 - acc: 0.3873 - val_loss: 1.3212 - val_acc: 0.4309\n",
      "\n",
      "Epoch 00023: val_acc did not improve\n",
      "Epoch 24/100\n",
      "2089/2089 [==============================] - 8s 4ms/step - loss: 1.2980 - acc: 0.4031 - val_loss: 1.3010 - val_acc: 0.4309\n",
      "\n",
      "Epoch 00024: val_acc did not improve\n",
      "Epoch 25/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.2949 - acc: 0.3887 - val_loss: 1.3095 - val_acc: 0.4228\n",
      "\n",
      "Epoch 00025: val_acc did not improve\n",
      "Epoch 26/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.2831 - acc: 0.3983 - val_loss: 1.3116 - val_acc: 0.4119\n",
      "\n",
      "Epoch 00026: val_acc did not improve\n",
      "Epoch 27/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.2957 - acc: 0.3925 - val_loss: 1.2962 - val_acc: 0.4255\n",
      "\n",
      "Epoch 00027: val_acc did not improve\n",
      "Epoch 28/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.2749 - acc: 0.4241 - val_loss: 1.3030 - val_acc: 0.4390\n",
      "\n",
      "Epoch 00028: val_acc did not improve\n",
      "Epoch 29/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.2922 - acc: 0.3992 - val_loss: 1.3176 - val_acc: 0.4282\n",
      "\n",
      "Epoch 00029: val_acc did not improve\n",
      "Epoch 30/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.2685 - acc: 0.4203 - val_loss: 1.3317 - val_acc: 0.4173\n",
      "\n",
      "Epoch 00030: val_acc did not improve\n",
      "Epoch 31/100\n",
      "2089/2089 [==============================] - 9s 4ms/step - loss: 1.2606 - acc: 0.4174 - val_loss: 1.3289 - val_acc: 0.4119\n",
      "\n",
      "Epoch 00031: val_acc did not improve\n",
      "Epoch 32/100\n",
      "2089/2089 [==============================] - 7s 4ms/step - loss: 1.2478 - acc: 0.4519 - val_loss: 1.3158 - val_acc: 0.4282\n",
      "\n",
      "Epoch 00032: val_acc did not improve\n",
      "Epoch 33/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.2457 - acc: 0.4251 - val_loss: 1.3097 - val_acc: 0.4472\n",
      "\n",
      "Epoch 00033: val_acc did not improve\n",
      "Epoch 34/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.2445 - acc: 0.4356 - val_loss: 1.3241 - val_acc: 0.4255\n",
      "\n",
      "Epoch 00034: val_acc did not improve\n",
      "Epoch 35/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.2435 - acc: 0.4323 - val_loss: 1.3209 - val_acc: 0.4390\n",
      "\n",
      "Epoch 00035: val_acc did not improve\n",
      "Epoch 36/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.2451 - acc: 0.4366 - val_loss: 1.3084 - val_acc: 0.4336\n",
      "\n",
      "Epoch 00036: val_acc did not improve\n",
      "Epoch 37/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.2294 - acc: 0.4562 - val_loss: 1.2993 - val_acc: 0.4390\n",
      "\n",
      "Epoch 00037: val_acc did not improve\n",
      "Epoch 38/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.2401 - acc: 0.4385 - val_loss: 1.3157 - val_acc: 0.4255\n",
      "\n",
      "Epoch 00038: val_acc did not improve\n",
      "Epoch 39/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.2015 - acc: 0.4663 - val_loss: 1.2948 - val_acc: 0.4526\n",
      "\n",
      "Epoch 00039: val_acc improved from 0.44986 to 0.45257, saving model to GRU_model.hdf5\n",
      "Epoch 40/100\n",
      "2089/2089 [==============================] - 8s 4ms/step - loss: 1.2132 - acc: 0.4466 - val_loss: 1.2956 - val_acc: 0.4282\n",
      "\n",
      "Epoch 00040: val_acc did not improve\n",
      "Epoch 41/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1986 - acc: 0.4696 - val_loss: 1.3050 - val_acc: 0.4363\n",
      "\n",
      "Epoch 00041: val_acc did not improve\n",
      "Epoch 42/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.2089 - acc: 0.4672 - val_loss: 1.3091 - val_acc: 0.4472\n",
      "\n",
      "Epoch 00042: val_acc did not improve\n",
      "Epoch 43/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1994 - acc: 0.4658 - val_loss: 1.3075 - val_acc: 0.4309\n",
      "\n",
      "Epoch 00043: val_acc did not improve\n",
      "Epoch 44/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1977 - acc: 0.4624 - val_loss: 1.2906 - val_acc: 0.4255\n",
      "\n",
      "Epoch 00044: val_acc did not improve\n",
      "Epoch 45/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1955 - acc: 0.4600 - val_loss: 1.2710 - val_acc: 0.4282\n",
      "\n",
      "Epoch 00045: val_acc did not improve\n",
      "Epoch 46/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.1877 - acc: 0.4686 - val_loss: 1.2843 - val_acc: 0.4255\n",
      "\n",
      "Epoch 00046: val_acc did not improve\n",
      "Epoch 47/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.1802 - acc: 0.4758 - val_loss: 1.2751 - val_acc: 0.4444\n",
      "\n",
      "Epoch 00047: val_acc did not improve\n",
      "Epoch 48/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1716 - acc: 0.4969 - val_loss: 1.2784 - val_acc: 0.4417\n",
      "\n",
      "Epoch 00048: val_acc did not improve\n",
      "Epoch 49/100\n",
      "2089/2089 [==============================] - 8s 4ms/step - loss: 1.1709 - acc: 0.4744 - val_loss: 1.2890 - val_acc: 0.4499\n",
      "\n",
      "Epoch 00049: val_acc did not improve\n",
      "Epoch 50/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.1562 - acc: 0.4931 - val_loss: 1.2825 - val_acc: 0.4472\n",
      "\n",
      "Epoch 00050: val_acc did not improve\n",
      "Epoch 51/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.1675 - acc: 0.4921 - val_loss: 1.2894 - val_acc: 0.4282\n",
      "\n",
      "Epoch 00051: val_acc did not improve\n",
      "Epoch 52/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1594 - acc: 0.5012 - val_loss: 1.2664 - val_acc: 0.4499\n",
      "\n",
      "Epoch 00052: val_acc did not improve\n",
      "Epoch 53/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1659 - acc: 0.4955 - val_loss: 1.2714 - val_acc: 0.4444\n",
      "\n",
      "Epoch 00053: val_acc did not improve\n",
      "Epoch 54/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1535 - acc: 0.4940 - val_loss: 1.2778 - val_acc: 0.4336\n",
      "\n",
      "Epoch 00054: val_acc did not improve\n",
      "Epoch 55/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1346 - acc: 0.5141 - val_loss: 1.2722 - val_acc: 0.4526\n",
      "\n",
      "Epoch 00055: val_acc did not improve\n",
      "Epoch 56/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1111 - acc: 0.5304 - val_loss: 1.2790 - val_acc: 0.4309\n",
      "\n",
      "Epoch 00056: val_acc did not improve\n",
      "Epoch 57/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1199 - acc: 0.5136 - val_loss: 1.2886 - val_acc: 0.4336\n",
      "\n",
      "Epoch 00057: val_acc did not improve\n",
      "Epoch 58/100\n",
      "2089/2089 [==============================] - 8s 4ms/step - loss: 1.1275 - acc: 0.5060 - val_loss: 1.2879 - val_acc: 0.4336\n",
      "\n",
      "Epoch 00058: val_acc did not improve\n",
      "Epoch 59/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.1081 - acc: 0.5156 - val_loss: 1.2737 - val_acc: 0.4363\n",
      "\n",
      "Epoch 00059: val_acc did not improve\n",
      "Epoch 60/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1181 - acc: 0.5127 - val_loss: 1.2871 - val_acc: 0.4282\n",
      "\n",
      "Epoch 00060: val_acc did not improve\n",
      "Epoch 61/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.1057 - acc: 0.5270 - val_loss: 1.2905 - val_acc: 0.4472\n",
      "\n",
      "Epoch 00061: val_acc did not improve\n",
      "Epoch 62/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0923 - acc: 0.5304 - val_loss: 1.3198 - val_acc: 0.4309\n",
      "\n",
      "Epoch 00062: val_acc did not improve\n",
      "Epoch 63/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0878 - acc: 0.5318 - val_loss: 1.3166 - val_acc: 0.4228\n",
      "\n",
      "Epoch 00063: val_acc did not improve\n",
      "Epoch 64/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0708 - acc: 0.5428 - val_loss: 1.3198 - val_acc: 0.4282\n",
      "\n",
      "Epoch 00064: val_acc did not improve\n",
      "Epoch 65/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0903 - acc: 0.5266 - val_loss: 1.3079 - val_acc: 0.4363\n",
      "\n",
      "Epoch 00065: val_acc did not improve\n",
      "Epoch 66/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0801 - acc: 0.5400 - val_loss: 1.2947 - val_acc: 0.4309\n",
      "\n",
      "Epoch 00066: val_acc did not improve\n",
      "Epoch 67/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 1.0652 - acc: 0.5414 - val_loss: 1.2969 - val_acc: 0.4255\n",
      "\n",
      "Epoch 00067: val_acc did not improve\n",
      "Epoch 68/100\n",
      "2089/2089 [==============================] - 8s 4ms/step - loss: 1.0566 - acc: 0.5472 - val_loss: 1.3008 - val_acc: 0.4309\n",
      "\n",
      "Epoch 00068: val_acc did not improve\n",
      "Epoch 69/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0634 - acc: 0.5395 - val_loss: 1.3211 - val_acc: 0.4201\n",
      "\n",
      "Epoch 00069: val_acc did not improve\n",
      "Epoch 70/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0369 - acc: 0.5668 - val_loss: 1.3198 - val_acc: 0.4363\n",
      "\n",
      "Epoch 00070: val_acc did not improve\n",
      "Epoch 71/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0501 - acc: 0.5620 - val_loss: 1.3125 - val_acc: 0.4092\n",
      "\n",
      "Epoch 00071: val_acc did not improve\n",
      "Epoch 72/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0325 - acc: 0.5586 - val_loss: 1.3193 - val_acc: 0.4255\n",
      "\n",
      "Epoch 00072: val_acc did not improve\n",
      "Epoch 73/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0114 - acc: 0.5778 - val_loss: 1.3179 - val_acc: 0.4146\n",
      "\n",
      "Epoch 00073: val_acc did not improve\n",
      "Epoch 74/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 0.9908 - acc: 0.5931 - val_loss: 1.3173 - val_acc: 0.4390\n",
      "\n",
      "Epoch 00074: val_acc did not improve\n",
      "Epoch 75/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 1.0048 - acc: 0.5744 - val_loss: 1.3154 - val_acc: 0.4255\n",
      "\n",
      "Epoch 00075: val_acc did not improve\n",
      "Epoch 76/100\n",
      "2089/2089 [==============================] - 7s 3ms/step - loss: 0.9975 - acc: 0.5831 - val_loss: 1.3170 - val_acc: 0.4336\n",
      "\n",
      "Epoch 00076: val_acc did not improve\n",
      "Epoch 77/100\n",
      "2089/2089 [==============================] - 8s 4ms/step - loss: 1.0085 - acc: 0.5797 - val_loss: 1.3157 - val_acc: 0.4390\n",
      "\n",
      "Epoch 00077: val_acc did not improve\n",
      "Epoch 78/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 0.9785 - acc: 0.5917 - val_loss: 1.3155 - val_acc: 0.4255\n",
      "\n",
      "Epoch 00078: val_acc did not improve\n",
      "Epoch 79/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 0.9746 - acc: 0.5936 - val_loss: 1.3076 - val_acc: 0.4472\n",
      "\n",
      "Epoch 00079: val_acc did not improve\n",
      "Epoch 80/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 0.9686 - acc: 0.5917 - val_loss: 1.3190 - val_acc: 0.4472\n",
      "\n",
      "Epoch 00080: val_acc did not improve\n",
      "Epoch 81/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 0.9638 - acc: 0.6017 - val_loss: 1.3058 - val_acc: 0.4390\n",
      "\n",
      "Epoch 00081: val_acc did not improve\n",
      "Epoch 82/100\n",
      "2089/2089 [==============================] - 6s 3ms/step - loss: 0.9594 - acc: 0.6060 - val_loss: 1.3073 - val_acc: 0.4390\n",
      "\n",
      "Epoch 00082: val_acc did not improve\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train_sub[:,:seq_len,:], y_train,\n",
    "                    batch_size=batch_size, epochs=num_epoch, shuffle=True,\n",
    "                    validation_split = 0.15, callbacks=callbacks_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_loss', 'val_acc', 'loss', 'acc'])\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f0130471860>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzt3X+cXHV97/HXZzeTZEJCNiQRkg2a1NqAQkgkKhrqRagEQSBFDCi1LdcWvfAowoMbCe0tIlWJpl6RqkWKVG25QIQQQbSxEpCCICYk/CZFBMxuJFlCNgLZkM3u5/5xzkxmZ8+ZHzs/zuzM+/l47GN3zpw557Pz43zm+9vcHREREYC2pAMQEZHGoaQgIiJZSgoiIpKlpCAiIllKCiIikqWkICIiWUoKIiUys++a2RdK3PcFM/uTSo8jUm9KCiIikqWkICIiWUoK0lTCaptlZvaYmb1uZt8xs4PN7Cdm9qqZ/czMpuTsf5qZPWlmvWZ2r5kdnnPfAjN7JHzcLcD4vHN92Mw2hY/9hZnNG2HMf21mvzazV8zsDjObGW43M/uamW03s13h/3REeN/JZvZUGFu3mf3vET1hInmUFKQZfQT4IPBHwKnAT4C/BaYRvOcvBDCzPwJuAi4CpgM/Bu40s7FmNhZYA/wbcBDwg/C4hI99J3AD8ClgKvBt4A4zG1dOoGZ2PHAVsBSYAbwI3BzefSLw/vD/6ADOAnaE930H+JS7TwKOANaVc16ROEoK0oz+yd23uXs38F/AL919o7u/AdwOLAj3Owu4y93/0937gX8E0sD7gGOAFHC1u/e7+63Ar3LO8dfAt939l+4+4O7fA94IH1eOc4Ab3P2RML7LgPea2WygH5gEHAaYuz/t7r8LH9cPvN3MDnT3ne7+SJnnFYmkpCDNaFvO330RtyeGf88k+GYOgLsPAluAzvC+bh86Y+SLOX+/BbgkrDrqNbNe4NDwceXIj+E1gtJAp7uvA74BfBPYZmbXmdmB4a4fAU4GXjSzn5vZe8s8r0gkJQVpZVsJLu5AUIdPcGHvBn4HdIbbMt6c8/cW4Ivu3pHzM8Hdb6owhgMIqqO6Adz9Gnc/GngHQTXSsnD7r9z9dOBNBNVcq8o8r0gkJQVpZauAU8zsBDNLAZcQVAH9AngQ2AdcaGZjzOwM4N05j/0X4NNm9p6wQfgAMzvFzCaVGcP/A841s/lhe8SXCKq7XjCzd4XHTwGvA3uAgbDN4xwzmxxWe/0eGKjgeRDJUlKQluXum4E/A/4JeJmgUfpUd9/r7nuBM4C/BHYStD+sznnseoJ2hW+E9/863LfcGO4G/h64jaB08lbg7PDuAwmSz06CKqYdBO0eAJ8AXjCz3wOfDv8PkYqZFtkREZEMlRRERCRLSUFERLKUFEREJEtJQUREssYkHUC5pk2b5rNnz046DBGRUWXDhg0vu/v0YvuNuqQwe/Zs1q9fn3QYIiKjipm9WHwvVR+JiEgOJQUREclSUhARkaxR16YQpb+/n66uLvbs2ZN0KDU3fvx4Zs2aRSqVSjoUEWlCTZEUurq6mDRpErNnz2bopJbNxd3ZsWMHXV1dzJkzJ+lwRKQJNUVS2LNnT8GEsHP3Xrbt2sPegUHGtrdx8OTxTJkwts5RVs7MmDp1Kj09PUmHIiJNqimSAlAwIXTv7GMwnPhv78Ag3Tv7AEZtYhARqZWmb2jetmtPNiFkDLqzbVfztz+IiJSr6ZPC3oHBsraPRG9vL9/61rfKftzJJ59Mb29v1eIQEalU0yeFse3D/8V7N2/nr763njnL72LRinWs2dhd0TniksLAQOHFsH784x/T0dFR0blFRKqpadoU4hw8efyQNoV7N2/nm/c8xxv7gpJCd28fl61+HIAlCzpHdI7ly5fz3HPPMX/+fFKpFBMnTmTGjBls2rSJp556iiVLlrBlyxb27NnDZz7zGc477zxg/5Qdr732Gh/60Ic49thj+cUvfkFnZyc//OEPSafTVXgGRERK1/QlhSkTxtI5JZ0tMfz7Q7/NJoSMvv4BVq7dPOJzrFixgre+9a1s2rSJlStX8vDDD/PFL36Rp556CoAbbriBDRs2sH79eq655hp27Ngx7BjPPvssF1xwAU8++SQdHR3cdtttI45HRGSkmr6kAEFiyPQ06nn1jch9tvb2Ve187373u4eMI7jmmmu4/fbbAdiyZQvPPvssU6dOHfKYOXPmMH/+fACOPvpoXnjhharFIyJSqqYvKeSb2RFdJRO3fSQOOOCA7N/33nsvP/vZz3jwwQd59NFHWbBgQeTI63HjxmX/bm9vZ9++fVWLR0SkVC2XFJYtnks61T5kWzrVzrLFc0d8zEmTJvHqq69G3rdr1y6mTJnChAkTeOaZZ3jooYdGfB4RkVprieqjXJnG5JVrN7O1t4+ZHWmWLZ474kZmgKlTp7Jo0SKOOOII0uk0Bx98cPa+k046iWuvvZZ58+Yxd+5cjjnmmIr/BxGRWjHPG9jV6BYuXOj5i+w8/fTTHH744QlFVH+t9v+KSOXMbIO7Lyy2X8tVH4mISDwlBRERyVJSEBGRLCUFERHJUlIQEZEsJQUREclSUkjAxIkTkw5BRCRSayaFx1bB146AKzqC34+tSjoiEZGG0HIjmnlsFdx5IfSHE+Dt2hLcBpi3dESHvPTSS3nLW97C+eefD8AVV1yBmXHfffexc+dO+vv7+cIXvsDpp59ejf9ARKRmalZSMLMbzGy7mT0Rc/9kM7vTzB41syfN7NxaxTLE3VfuTwgZ/X3B9hE6++yzueWWW7K3V61axbnnnsvtt9/OI488wj333MMll1zCaBs9LiKtp5Ylhe8C3wC+H3P/BcBT7n6qmU0HNpvZje6+t4Yxwa6u8raXYMGCBWzfvp2tW7fS09PDlClTmDFjBhdffDH33XcfbW1tdHd3s23bNg455JARn0dEpNZqlhTc/T4zm11oF2CSmRkwEXgFqP180ZNnBVVGUdsrcOaZZ3Lrrbfy0ksvcfbZZ3PjjTfS09PDhg0bSKVSzJ49O3LKbBGRRpJkQ/M3gMOBrcDjwGfcfTBqRzM7z8zWm9n6np6eys56wuWQyls7IZUOtlfg7LPP5uabb+bWW2/lzDPPZNeuXbzpTW8ilUpxzz338OKLL1Z0fBGRekgyKSwGNgEzgfnAN8zswKgd3f06d1/o7gunT59e2VnnLYVTr4HJhwIW/D71mhE3Mme84x3v4NVXX6Wzs5MZM2ZwzjnnsH79ehYuXMiNN97IYYcdVlncIiJ1kGTvo3OBFR60vv7azJ4HDgMervmZ5y2tOAlEefzxx7N/T5s2jQcffDByv9dee63q5xYRqYYkSwq/BU4AMLODgbnAbxKMR0Sk5dWspGBmNwHHAdPMrAv4HJACcPdrgX8AvmtmjwMGXOruL9cqHhERKa6WvY8+VuT+rcCJVTwfQUem5qaxDiJSS00xzcX48ePZsWNH018w3Z0dO3Ywfvz4pEMRkSbVFNNczJo1i66uLirurjoKjB8/nlmzKhtTISISpymSQiqVYs6cOUmHISIy6jVF9ZGIiFSHkoKIiGQpKYiISJaSgoiIZCkpiIhIlpKCiIhkKSmIiEhWU4xTKNeajd2sXLuZrb19zOxIs2zxXJYs6Ew6LBGRxLVcUlizsZvLVj9OX/8AAN29fVy2OpjyWolBRFpdy1UfrVy7OZsQMvr6B7hk1aPMWX4Xi1asY83G7oSiExFJVsuVFLb29kVuHwgn01PJQURaWcuVFGZ2pIvu09c/wMq1m+sQjYhIY2m5pLBs8VzSqfai+8WVKEREmlnLVR9lqoQyvY/azLJVR7lKKVGIiDSblksKECSGTHLI740EkE61s2zx3KTCExFJTEsmhVz5JQeNWxCRVtbySQGGlhxERFpZyzU0i4hIPCUFERHJUlIQEZEsJQUREclSUhARkSwlBRERyVJSEBGRLCUFERHJUlIQEZEsJQUREclSUhARkSwlBRERyVJSEBGRrJolBTO7wcy2m9kTBfY5zsw2mdmTZvbzWsUiIiKlqWVJ4bvASXF3mlkH8C3gNHd/B/DRGsYiIiIlqFlScPf7gFcK7PJxYLW7/zbcf3utYhERkdIk2abwR8AUM7vXzDaY2Z/H7Whm55nZejNb39PTU8cQRURaS5Irr40BjgZOANLAg2b2kLv/d/6O7n4dcB3AwoULvZ5BrtnYraU6RaRlJJkUuoCX3f114HUzuw84ChiWFJKyZmM3l61+nL7+AQC6e/u4bPXjAEoMItKUkqw++iHwx2Y2xswmAO8Bnk4wnqw1G7tZtGIdF92yKZsQMvr6B1i5dnNCkYmI1FbNSgpmdhNwHDDNzLqAzwEpAHe/1t2fNrP/AB4DBoHr3T22+2q95JcOomzt7atjRCIi9VOzpODuHythn5XAylrFMBIr124umBAAZnak6xSNiEh9aURznmKlgHSqnWWL59YpGhGR+lJSyFOoFNDZkeYjR3eycu1m5iy/i0Ur1rFmY3cdoxMRqS0lhTzLFs8lnWofsi2daufqs+azbPFcbtvQTXdvH87+3khKDCLSLJQU8ixZ0MlVZxxJZ0caIygdXHXGkSxZ0BnZ3qDeSCLSTJIcp9CwlizojByHENfeoN5IItIsVFIoQ1x7g3ojiUizUFIoQ1R7Q6rN2L13nxqeRaQpqPqoDJkqpcxcSJPTKV7fu4+du/sBTYMhIqOfSgplWrKgkweWH8/zK07hgHFj6B8YOj+fGp5FZDRTUqiAGp5FpNkoKVRADc8i0myUFCoQN9BN02CIyGilhuYK5Dc8z+xI84HDprNy7WYuvmWTFuURkVHH3Ou6kFnFFi5c6OvXr086jEhR024b4AQjo5UgRCQpZrbB3RcW20/VR1UUNQ1GJuVqniQRGQ2UFKqoWK8jdVcVkUanpFBFpfQ6UndVEWlkSgpVFNUbKZ+6q4pII1PvoyrK7Y3U3duXbWTOUHdVEWl06n1UQ2s2dg/rrnrPMz3Z2+qNJCL1UmrvI5UUaih3XYb87qqaPE9EGpHaFOpEq7aJyGigpFAnmjxPREYDJYU60eR5IjIalJQUzOwzZnagBb5jZo+Y2Ym1Dq6ZaPI8ERkNSi0p/E93/z1wIjAdOBdYUbOomtCSBZ1cdcaRdHakMYK5kK4640g1MotIQym195GFv08G/tXdHzUzK/SAhvLYKrj7StjVBZNnwQmXw7yldQ8jtzeSiEgjKjUpbDCznwJzgMvMbBIwWLuwquixVXDnhdAfNuju2hLchkQSQznyxzloXIOI1FqpSeGTwHzgN+6+28wOIqhCanx3X7k/IWT09wXb650U8kosv3rr33DRU2+LvOhrXIOIJKGkEc1mtgjY5O6vm9mfAe8Evu7uL9Y6wHxlj2i+ooOhk01kGFzRW62wissvsQB9PpZL+/+KOwaPBSDVZkwcP4be3f20mTEQ8dp0dqR5YPnxdQtbRJpDtddT+Gdgt5kdBXwWeBH4fgXx1c/kWeVtr5WIEkva9vLZMauyt/sHnZ27+3GITAigcQ0iUlulJoV9HhQpTicoIXwdmFS7sKrohMshlTcWIJUOttfTrq7IzTNtR1mH0bgGEamlUpPCq2Z2GfAJ4C4zawdStQuriuYthVOvgcmHAhb8PvWawu0Jj62Crx0RVD197YjgdqViSiZbfWrJhzCCtoVFK9ZpBTcRqYlSk8JZwBsE4xVeAjqBlYUeYGY3mNl2M3uiyH7vMrMBMzuzxFjKN28pXPxE0IZw8RPFE8KdFwa9lPD9vZUqTQwRJZY+H8tX9hVu7G4Pe/7mTsOtpT1FpFZKSgphIrgRmGxmHwb2uHuxNoXvAicV2iEscXwZWFtKHHVRqLcSjLwUEVFieeLoL7DhwA9iQEc6Rap96NCPdKqdry49is6O9LCm8g8O/Jxjfvg/qluaEZGWV2rvo6UEJYN7Cb60/jGwzN1vLfK42cCP3P2ImPsvAvqBd4X7FTwe1GE9hdjeSkD6INj7Ggzs3b8tlS5eHVWiuHEJc5bfNSSi09ruZ0XqeiZYbeIQkeZT7fUU/g54l7tvDw8+HfgZUPQiXiDATuBPgeMJkkKhfc8DzgN485vfPNJTlmbyrLDqKELfK8O3VTLmIW/cwpITLmfJ8uHHmdmRpjun19Fnx6wamhAycdz+aVh9XqKjtkVkdCu1TaEtkxBCO8p4bJyrgUvdfaDYju5+nbsvdPeF06dPr/C0RUT1ViompmdRQWW0XeRPpjfTXo4+pg9EHmvNxm4WrVjHnOV3qZFaRAoqtaTwH2a2FrgpvH0W8OMKz70QuDmcQmkacLKZ7XP3NRUetzKZb9d3XxlfYsg3kjEPZYy0zl37eWtvHy/ZNGYSkxjyjrVmYJFGRotIyUptaF4GXAfMA44CrnP3Sys5sbvPcffZ7j6boBrq/MQTQkamt9LkQ4vvmz/modSG6LjSRcz2Je0P8MC4C3l+/DlM4A32egn5fFeXVnwTkbKUvEazu98G3Fbq/mZ2E3AcMM3MuoDPEY5tcPdrywszISdcPmxqCtpSMG4S9O0cXndfzuR7cW0XUaWOvON28Cpv0M4rPpEOXmcQY4xFzE84eRZbt/VxWtv9fHbMKmbay2z1aXxl31Lu7D22zCdDRFpBwaRgZq8SO3EQ7u4Hxj3W3T9WahDu/pel7ltXQ6qSSph2u5zJ96ISTtxI64jjjrMBegbH886910X2RnrD29m9q5fnxn0cgLawt+sse5kVqes5KDUWOKXYMyAiLaZgUnD30TGVRbnKWV9h3tLSe/GUUyVUTsIpMkXGHYPHQj9haWAHO/0AJtkepvDq/pUwckywvVw+cA1c8fXgvG87EZ79aeLrTYhI8kquPmoatVxfoZwqocz5SjlnzHG327TsSOc7Bo/ljr1BldD9Yy9kqr1W8JBtHlY37doC67+z/45RtN6EiFRfpd1KR59iI5bLlduwvPd1aB+bt4MFF9pKRh3HTOp3yBlf4vkVp9CZN0lebJfVUlXyfIjIqNZ6SaHMXj8F5Y816HsF3IORz8CQGYtGModSJuGsPg/GpMPjDp/UL38cw1afVv7/km/XFk2hIdKCWi8pxFXlWFv5F8GoUsdgP4w9IOzOmtdGX8438KiEs68Pzrhu2KR+SxZ0ctUZR9LZkcaAb7Z9nN0+tMQy6GE01k7pqjghoIiMCq2XFOJGLMeMBi6oUKmj0hJJmdVcSxZ08sDy43l+xSlcdcU/8OTRX+AlpjPoxktMZ8PRX8Gu2AV/em35I7ZVnSTSMlqvoTm/14+1hQkhR6nzGRVrWC6n0TlfhUnlXad9Ck77FACHhD9AdK+nnN5Hjkd1WMJ3dUVur5n8HmLqISVSF62XFGBor58rOqL3KeXiW2ysQanjEKKU25OpHAV6PW274g85hJ7h25m2P7HUWlQPsXJ6SJXT5VhEhmjNpJCrkotvKWMNyrk45V7M0lOCnkz503RXYRnRuCm6Aa7a+1GuyhsIt9vH8qX+j7JhxTqufvuzvOu5f6rtBTeq6ixfXGkuKqGsOR9+cmn0KHRJTqHkXc3EXk6pM6kvFA30Raak9RQaSdXXU8i/iEAyaxNExVFoSo0RWrOxe8gEeQCpNmPi+DH07u6nzYxT7L+yA+G2+lS+sm8pdwwey2lt9/Pl1PWka7WOQ/aDUeJEhEDQGyvnufnaEcUfXyjmBvpwNrWo93umt1411y2JPE+ezLGh+GcwN6GkpwT7RN1Xyu3Me6tO16BS11NQUoDGuBDEXcwmHxr0NqqSRSvWDVmboZjceZMGaYuZYyknxvzSDpSW1Er58BaS/fBGrHkRJep5rfTD2Qjvo9GilOSdbySfhXLOY+3D2xdrKfPeivsiVOXPvpLCaBO74psFa0tXSf4qbnHawxLDsBXeIoUxFruwF7rAjuQiUbESSxmFPpxDSje5K2lTk5Je0yi0wmFBVt5zOeLz1EnRRGRV62hR7ZXXpNZq2bCcI38VtziD7vztuB8wgWIJAQbNaLuiI7onV678doDcb9aFPriTDx36oajah9zz2hxiShlxnQ6GJcG8uAb79x9T04cMVWiFw4K8+HOZ+74q9p5MWtHYvO5T0aik0CjqVK8Y1aYQpbMjzf17zsCKXIDdwcroq+rhY35vk5hke2j3/sIPiPqWXk6pIqp+ulzWDj44/BvbSC44Va4SGLUqrS6E/a9LbjVlekrlr/doUeZ7qdSSQusNXmtU85YGCWDyoURNZVEt+aOfO9IpUu1Dr+rpVDvLFs/FYkop+7yNQTf2eVtZCQGCCpY2C9aEKJYQ9rWPj+5tVeqSqZMPhUufh9O/uf95HYncgY3rv7N/lPlIvoG28vQhufOE3X0lHPXxnIWs8l6bttT+aV3iZF6XvlfCEln4d1RCsHayn6uFnyy8gFZm3/RBEXOZNZCRTM1TApUUJL6LasS3ud0+luX9f8Udg8fym3Efz67TUC3u4BhbfSpXczbH/un50cuG5jdol9pbpZEaHispCdarO2ex85bavbPYa1Qo5qq0N0W0zZVSOi/0/1bS+yiulJlJVqX8vzUqKSgpSGE5H4rd6UO4/PWPcOve9wHBFN2z2obPyLrP22jD2ekHYAYdvI7hJZUqugancezea7K3OzvSPLD8+LLiLHghLLXaYvKhI2y/KNCtMu48UT23yv0fhvS+ymvwrlY1ZKXdO6OUcmGrRlVT3HmSHJcQl5Cg9Oe5jFjV0CzVkTP6eQJw7MZuHgxLFf84cBZfsn8ZNtAtU5LIdf/YC5lVZErv3T6Wr+wb+ibfWmr32VLXpsgfcBj3DfaEy0sfM5Hb5hD3TTkuuWSqAIqt81Gs8TS3UTtuIsZC3YFLKXWU0oaSO09WKRfxUqpASpmappBCgz7LWUSrmsod+FrHaV5UUpARW7Oxm/tv/xYXcfOwgW75YpcMtQlM9tf4HVP5cv/wx5ZcUqhE3EWxnG/GxT6gcVUgmYRS6EJXjcby3OqTYtU65ZZ2KjGShvdir4u6AkdS9ZHURX57xAcOm849z/REdnvdPxBuBy/ZVFbuO4vb9y2KPXamEqQzbyqOuqrWxHzVqAKpRG5PnUbpnVOtNpVyBkm2MCUFSVTcyOnMN/+4+9vNGHDPrxUnnWrnqjOOTCYxVMto6j9fC/oGnyglBUlU1HiI3At73MhqI36AXbsZg+7DJvEblaox0rbgt/8wrdaiB1XkuI1CbS9ljkKWmlBDsyQqc8GOm4017sI/syMd27g8EH6B6e7t47LVjw85z6gTN6K31It4sa6TmQtw3NTwI1VuV18N1ht1VFKQRBQqSaxcu7mkqThGdcmh3C6JI616KaWPf6HurKWet1FmG5ZYKilIQytWkihlKo7cksOyHzzK5+98kt7d/UxOpzCD3t39Qxq/8xvDo85bN9VeiyNO1EJQhS70I+23X8r/I6OCSgrSkHJ7NbWFjc+1ktvWUWgBolFLU3oLamiWJlLqJH6VyHR7LdQ4LjKaaUI8aRr5k/i1lzsLXwm6e/u46JZNwxJPX/8AK9durvr5RBqV2hRkVFiyoDP7bb0eJYdcJU+1IdIElBRk1MlvpJ6cTvH63n30D9SmKnRmRwnTdIs0CSUFGZVySw4wtGG6nN5HxdJIZm0JkVahpCBNIT9JlCpuug2InnOpUO+kpuy5JC1HSUFaWjk9jvLbMnJHVgOx9ykxyGhSsy6pZnYD8GFgu7sfEXH/OcCl4c3XgP/l7o8WO666pEq1lfrtP268RGfY5tD08zXJqJb4OAUzez/Bxf77MUnhfcDT7r7TzD4EXOHu7yl2XCUFqZdq93JKtRkTx4+JbetQ0pBaSnyaC3e/z8xmF7j/Fzk3HwKiV4kXScjKtZur2u21f9DZubsfCEoV//7Qb7P3qbpJGkWjDF77JPCTuDvN7DwzW29m63t6euoYlrSyeo9P6Osf4JJVjzJn+V0sWrGONRu763p+EWiApGBmHyBICpfG7ePu17n7QndfOH369PoFJy0tbnxCuxnFxlSPdNT1gDvO/pKDEoPUW6JJwczmAdcDp7v7jiRjEcm3bPFc0qn2IdvSqXa+uvQonl9xSraBOV9nR5qvLj1q2GPL1dc/wEW3bFKpQeoqsS6pZvZmYDXwCXf/76TiEIlTbHrvuO6suftUY9R1/tTgaqSWWqpl76ObgOOAacA24HNACsDdrzWz64GPAC+GD9lXSsu4eh9JIylnwFr+vrkX9kqnB2/66b+lYol3Sa0VJQVpRtXo/lrq9N9KGq1JSUFklMlcrEtZijSKEb/2dWYQXVQ1ltaMaA1KCiKj1EhLDe0VVEFp5HXzS3zwmoiMzEgbqStpk8hd71qD6FqbkoJIAyo0NXi1G6nzZVabU1JoTUoKIqNAoanB5yy/K/ZxI61S0mpzrSvxEc0iUpm4kddxg+hSbcaUCamC6107aNBci1JSEBnl4kZeZxqMrzrjSDo70hhBolj50aPYePmJPL/ilIIjrzXVRmtS9ZHIKFds5HWhqqfcx0Z1ZS3WvqAxD81HXVJFBAjaJqKuBgY8v+IUYPha2BrzMHqU2iVV1UciAsS3TWS2Z8ZPdPf24UBvX/+wbrKZkkVm/0Ur1mkq8FFG1UciAkRP8JdqM3bv3cec5XeV3PV1a29fwfWsVYpobKo+EpGsYtVDpYrrCquR08nRiGYRKVtuo/SiFevo7esf0XHiShQaOd341KYgIpFKGcCWGfMwErntD9I4lBREJFKx5UhzxzyMbPHRoMRQSkO0Gq3rR20KIhIparbWuC6ni1asKzhldymN1EYwkrozr72hnDgknrqkikhFokZDx12Ii61nXcqa1ZmUkT+SeuXazcOmEVfVU+2opCAiVVFsdHPu/aVcdTo70jyw/PjYQXWwf2Eh9WQqTr2PRKSuCk2nkX9/XHVTrkxDd9xqchCULtSTqbpUfSQidRdV3ZSvzYw5y+/i9Tf2kWov3JSt6qTqUVIQkbrLba8AInsvDbhnp9PAyU73Hae7ty+yZ5J6LpVHbQoikrjc9oa4nkqZNoZiVU+pNmPi+DH07u7XpH051PtIREaNJQs6eWD58Ty/4hQGY76oZtoYilU99Q86O3f3lzRpnwynhmYRaShxDcuZwXTF1oAoRSbBaD2I4VQGXrL5AAAJQklEQVRSEJGGUmgluYxMyaIzZtR1MTM70sOmAtdKcwElBRFpKJUOmismk2A0KC6aqo9EpOEUG/OQux9QcLrv3IbnmR1pPnDY9IJVT6VMBNjM1PtIRJpKoXaCqHmU8jXrmg8a0SwiLalQKSOqyihfq6/5oDYFEWkZhaqG2m340LhWbGNQUhCRlhG3RkRnR7ro+IhWoaQgIi2jUHfXuISRmYOpVabIUJuCiLSM/N5K+Y3JUY3QcW0M5UwVPpoarWuWFMzsBuDDwHZ3PyLifgO+DpwM7Ab+0t0fqVU8IiIQ3xCdnzCi5mDKbWPITSBRCaPQ/Y2sZl1Szez9wGvA92OSwsnA3xAkhfcAX3f39xQ7rrqkikg9xC3uk1nYZyTLj2Ym9UtC4hPiuft9wCsFdjmdIGG4uz8EdJjZjFrFIyJSjkJtDHED3zLTfcetRz0aGq2TbGjuBLbk3O4Ktw1jZueZ2XozW9/T01OX4ESktcVNoRF3wS9FXKJpJEkmhaj1MiKfbXe/zt0XuvvC6dOn1zgsEZHhczBFjWMoR/6kfo0qyd5HXcChObdnAVsTikVEZJjcRuk5y++K3a89pg0hd8qMzJxLF9+yqaF7IyVZUrgD+HMLHAPscvffJRiPiEisQgPfvrr0qMjxD19dehTPrziFZYvnctuG7iHTdC/7waMsuPKnDTcGomZJwcxuAh4E5ppZl5l90sw+bWafDnf5MfAb4NfAvwDn1yoWEZFKFRr4Vmy676g5l3JXiOvu7ePiWzYxuwEShGZJFREp0UgHpMV1b42TP913NaqaSu2SqqQgIlJji1asG/HSoVCdJJH4OAUREQmMZIW4XPlVTbVcNlRJQUSkxvLbHDrSKVLtI+/iWsspvTUhnohIHeTPuZRpn+ju7cOIGaRVQK1GRyspiIgkIDdJ5DZgR60zHaVWo6OVFEREEhZXiohLErUcHa2kICLSYAoliVqPhlZSEBFpcHFrQNSCeh+JiEiWkoKIiGQpKYiISJaSgoiIZCkpiIhI1qibEM/MeoAXR/jwacDLVQynWhRX6RoxJmjMuBoxJmjMuBoxJqhuXG9x96JLV466pFAJM1tfyiyB9aa4SteIMUFjxtWIMUFjxtWIMUEycan6SEREspQUREQkq9WSwnVJBxBDcZWuEWOCxoyrEWOCxoyrEWOCBOJqqTYFEREprNVKCiIiUoCSgoiIZLVMUjCzk8xss5n92syWJxjHDWa23cyeyNl2kJn9p5k9G/6eUueYDjWze8zsaTN70sw+0yBxjTezh83s0TCuz4fb55jZL8O4bjGzsfWMK4yh3cw2mtmPGiimF8zscTPbZGbrw21Jv4YdZnarmT0Tvr/e2wAxzQ2fo8zP783sogaI6+Lwff6Emd0Uvv/r/r5qiaRgZu3AN4EPAW8HPmZmb08onO8CJ+VtWw7c7e5vA+4Ob9fTPuASdz8cOAa4IHx+ko7rDeB4dz8KmA+cZGbHAF8GvhbGtRP4ZJ3jAvgM8HTO7UaICeAD7j4/p2970q/h14H/cPfDgKMInrNEY3L3zeFzNB84GtgN3J5kXGbWCVwILHT3I4B24GySeF+5e9P/AO8F1ubcvgy4LMF4ZgNP5NzeDMwI/54BbE74+foh8MFGiguYADwCvIdghOeYqNe2TrHMIrhoHA/8CLCkYwrP+wIwLW9bYq8hcCDwPGGHlkaIKSLGE4EHko4L6AS2AAcRrHPzI2BxEu+rligpsP8Jz+gKtzWKg939dwDh7zclFYiZzQYWAL9shLjCappNwHbgP4HngF533xfuksRreTXwWWAwvD21AWKCYO33n5rZBjM7L9yW5Gv4B0AP8K9hVdv1ZnZAwjHlOxu4Kfw7sbjcvRv4R+C3wO+AXcAGEnhftUpSsIht6oubx8wmArcBF7n775OOB8DdBzwo5s8C3g0cHrVbveIxsw8D2919Q+7miF2TeH8tcvd3ElSTXmBm708ghlxjgHcC/+zuC4DXqX/1Vaywfv404AcNEMsU4HRgDjATOIDgdcxX8/dVqySFLuDQnNuzgK0JxRJlm5nNAAh/b693AGaWIkgIN7r76kaJK8Pde4F7Cdo8Oswss5RsvV/LRcBpZvYCcDNBFdLVCccEgLtvDX9vJ6gjfzfJvoZdQJe7/zK8fStBkmiU99WHgEfcfVt4O8m4/gR43t173L0fWA28jwTeV62SFH4FvC1syR9LUGS8I+GYct0B/EX4918Q1OnXjZkZ8B3gaXf/vw0U13Qz6wj/ThN8cJ4G7gHOTCIud7/M3We5+2yC99E6dz8nyZgAzOwAM5uU+ZugrvwJEnwN3f0lYIuZzQ03nQA8lWRMeT7G/qojSDau3wLHmNmE8POYea7q/75KqoGn3j/AycB/E9RJ/12CcdxEUGfYT/BN6pMEddJ3A8+Gvw+qc0zHEhRLHwM2hT8nN0Bc84CNYVxPAJeH2/8AeBj4NUHRf1xCr+VxwI8aIabw/I+GP09m3uMN8BrOB9aHr+EaYErSMYVxTQB2AJNztiX9XH0eeCZ8r/8bMC6J95WmuRARkaxWqT4SEZESKCmIiEiWkoKIiGQpKYiISJaSgoiIZCkpiNSRmR2XmVlVpBEpKYiISJaSgkgEM/uzcC2HTWb27XBivtfM7Ktm9oiZ3W1m08N955vZQ2b2mJndnpmH38z+0Mx+Fq4H8YiZvTU8/MScNQZuDEewijQEJQWRPGZ2OHAWwQRz84EB4ByCScoe8WDSuZ8Dnwsf8n3gUnefBzyes/1G4JserAfxPoKR7BDMQnsRwdoef0Awn5JIQxhTfBeRlnMCweIrvwq/xKcJJkcbBG4J9/l3YLWZTQY63P3n4fbvAT8I5yHqdPfbAdx9D0B4vIfdvSu8vYlgfY37a/9viRSnpCAynAHfc/fLhmw0+/u8/QrNEVOoSuiNnL8H0OdQGoiqj0SGuxs408zeBNl1jt9C8HnJzFj5ceB+d98F7DSzPw63fwL4uQfrUXSZ2ZLwGOPMbEJd/wuREdA3FJE87v6Umf0fglXM2ghmtL2AYJGYd5jZBoKVsc4KH/IXwLXhRf83wLnh9k8A3zazK8NjfLSO/4bIiGiWVJESmdlr7j4x6ThEaknVRyIikqWSgoiIZKmkICIiWUoKIiKSpaQgIiJZSgoiIpKlpCAiIln/H6eqFuQuOHKDAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f016ac9bc50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# list all data in history\n",
    "print(history.history.keys())\n",
    "\n",
    "# summarize history for accuracy\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', 'val'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# summarize history for loss\n",
    "plt.plot(history.history['loss'],'o')\n",
    "plt.plot(history.history['val_loss'],'o')\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'val'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\r",
      "100/100 [==============================] - 1s 7ms/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[1.1532464027404785, 0.50999999046325684]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# evaluate model on test set\n",
    "model = load_model('GRU_model.hdf5')\n",
    "model.evaluate(X_test_sub,y_test,batch_size=N_test)"
   ]
  },
  {
   "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}