part_2_neural_network.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import progressbar\n",
    "from schrodinger_solver import solver\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "L = 64"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def reducing(_in):\n",
    "    \"\"\"\n",
    "       A reducing convolutional layer has 64 filters of size 3x3.\n",
    "       We use stride 2 to half the data size.\n",
    "       We use ReLU activation\n",
    "    \"\"\"\n",
    "    return tf.contrib.layers.conv2d(_in, 64, kernel_size=3, stride=2, activation_fn=tf.nn.relu)\n",
    "\n",
    "def nonreducing(_in):\n",
    "    \"\"\"\n",
    "       A nonreducing convolutional layer has 16 filters of size 4x4.\n",
    "       We use stride 1 to preserve the data size.\n",
    "       We use ReLU activation.\n",
    "    \"\"\"\n",
    "    return tf.contrib.layers.conv2d(_in, 16, kernel_size=4, stride=1, activation_fn=tf.nn.relu)\n",
    "    \n",
    "def CNN(_in):\n",
    "    net = tf.reshape(_in, (-1, L, L, 1))\n",
    "    #If you're using 256x256 potentials, you'll want 6 modules.\n",
    "    #We'll use 4 since we're using 64x64 potentials\n",
    "    #  e.g. for 256x256 use   for moduleID in range(6):\n",
    "    for moduleID in range(4):\n",
    "        net = nonreducing(nonreducing(reducing(net)))\n",
    "    net = tf.reshape(net, (-1, 4*4*16))\n",
    "    net = tf.contrib.layers.fully_connected(net, 1024, activation_fn=tf.nn.relu)\n",
    "    net = tf.contrib.layers.fully_connected(net, 1, activation_fn=None)\n",
    "    return net"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#data comes in a [ batch * L * L * 1 ] tensor, and labels a [ batch * 1] tensor\n",
    "x = tf.placeholder(tf.float32, (None, L, L, 1), name='input_image')\n",
    "y = tf.placeholder(tf.float32, (None, 1))\n",
    "\n",
    "predicted = CNN(x)\n",
    "#define the loss function\n",
    "loss = tf.reduce_mean(tf.square(y-predicted))\n",
    "#create an optimizer, a training op, and an init op\n",
    "optimizer = tf.train.AdamOptimizer(learning_rate=0.00001)\n",
    "train_step = optimizer.minimize(loss)\n",
    "init = tf.global_variables_initializer()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sess = tf.InteractiveSession()\n",
    "sess.run(init) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "S = solver(limit=20, L=L, number=50000)\n",
    "data, labels = S.generate_file()\n",
    "train_data = data[:40000]\n",
    "test_data = data[40000:]\n",
    "train_labels = labels[:40000]\n",
    "test_labels = labels[40000:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|#############################################|loss: 1.12e-06|Time: 0:10:06\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'./chkpts/'"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "BATCH_SIZE = 1000\n",
    "EPOCHS = 100\n",
    "\n",
    "with progressbar.ProgressBar(max_value=EPOCHS,\n",
    "                             widgets=[progressbar.Percentage(),\n",
    "                                      progressbar.Bar(),\n",
    "                                      progressbar.DynamicMessage('loss'),'|',\n",
    "                                      progressbar.ETA()], \n",
    "                             redirect_stdout=False).start() as bar:\n",
    "    for epoch in range(EPOCHS):        \n",
    "        for batch in xrange(train_data.shape[0] / BATCH_SIZE):\n",
    "            _, loss_val = sess.run([train_step, loss],\n",
    "                       feed_dict={\n",
    "                            x: train_data[batch*BATCH_SIZE:(batch+1)*BATCH_SIZE],\n",
    "                            y: train_labels[batch*BATCH_SIZE:(batch+1)*BATCH_SIZE]\n",
    "                        }\n",
    "                      )\n",
    "            \n",
    "            bar.update(epoch, loss=float(loss_val))\n",
    "            \n",
    "saver = tf.train.Saver()\n",
    "saver.save(sess, save_path='./chkpts/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "BATCH_SIZE = 1000\n",
    "bar = progressbar.ProgressBar()\n",
    "prediction = []\n",
    "for batch in xrange(test_data.shape[0] / BATCH_SIZE):\n",
    "    batch_predictions = sess.run(predicted,\n",
    "                           feed_dict={\n",
    "                                x: test_data[batch*BATCH_SIZE:(batch+1)*BATCH_SIZE],\n",
    "                            }\n",
    "                          )  \n",
    "    prediction.append(batch_predictions.flatten())\n",
    "    \n",
    "prediction = np.array(prediction).flatten()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfsAAAHwCAYAAAChTMYRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4wLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvpW3flQAAIABJREFUeJzt3XucXVV58PHfwyQhiFwCQawJSNBg\nhaoEEby8glaUWwWptIKiaKmUCtW32r5itWJRWkVrtb5UQKFe0CJeaqMFERX0tYok3AVEQrgNQTSQ\nhCSEhMw87x97D+xzmclJMjvJ7PP7fj7nk73WXmvPWnMm59lrr3X2jsxEkiQ111abuwGSJKleBntJ\nkhrOYC9JUsMZ7CVJajiDvSRJDWewlySp4Qz22mwiIiPi2eX2uRHx95uhDVdFxJ+P8zE/FBEXjecx\nJWljGOy1ThFxd0SsiYjpbfk3lAF7j439GZl5SmZ+eGOPM9E09cQgIvaIiCsj4tGI+FVEHDJG2S+U\nf18rKq+BynGybd/fV+rOiIj/ioiHI2IwIk5pO/ZrI+KXZb2fRcTebfv/OiJ+ExHLIuLCiNi6su/D\nEXFzRKyNiA+11fu7tjatiojhkf8jPbQrI2Jlpf7nK/u2Lk9+HyzrfyciZlT2PzciflS2eUFEHNN2\n7D+NiNsiYnlE3BoRrxvzzVoP5WfBIW15b42In47Xz1A9DPbq1V3A8SOJiHgesM3ma47WV0RM6iVv\nfY8xiv8Argd2Bt4PfCMidhmj/NmZ+dTKa6ht/46VfdWTwoso/jZ3BY4E/jEiXlm2dTbwFeAUYEfg\nO8DckT5ExKHA6cCrgD2APYF/qBx7AfB/gP9ub2xm/mO1vcDHgKsyc/G62lXxgsoxqleX3gW8BHg+\n8AxgKfCZss2TgP8CvgvsBJwMXBQRe5X7Z5Q/+93A9sDfAl+NiKe190H9xWCvXn0ZeEslfSLwpWqB\nckTyiYi4txyVnBsR21T2/21EPBARiyLiz9rqfiEiPlJuT4uI70bE7yJiSbk9s1L2qnLU9T/l6OX7\n7VcdKmXHPFbpWRFxTTlS+q+I2KmsOzUiLoqIhyJiaUTMi4hdy33PiIi55chrQUS8fZSf/4qIGGzL\nuzsiDomIw4C/A95Qju5uLPfvEBEXlL+r+yPiI1GOdLscf6uIOD0i7izbeUml/SOj4pMi4l7gR93y\nyrJHRcQtZT+viojntrX3vRFxE7ByXQG/DDz7AWdk5qrM/CZwM/D6seqtr4h4KvAK4KzMfDwzbwS+\nAYz8bR0K/L/M/GlmrqUIyDOAg8v9JwIXZOYtmbkE+DDw1pHjZ+YXM/MyYPk62hHAm4Ev9tiudZkF\nXJ6ZD2bmY8DFwD7lvt+nOAH4l8wcyswfAf9T/nyAmcDSzLwsC/8NrASeNUrbvxAR/xYRl5V/g/8T\nEU+PiE+V/19+FRFzemz3yDFH/h5Hriwcs+5aqpvBXr26Gtg+ikuIA8AbKEYQVR8D9gL2BZ5N8cH6\nQYAysP0N8GpgNjDqZV2Kv8t/B54J7A6sAv5vW5k3Am8DngZMKY+9ocd6C8UH8TOAtcC/lvknAjsA\nu1GMUE8p60Mxch0s6xxLMXJ71Rh96pCZ3wP+EfhaObp7Qbnri2U7ng3MAV4DjLau4J3A6ygC2DOA\nJcA5bWUOBp5LEfw68srg/B/A/wZ2AS4FvhMRUyrlj6cYoe6YmWvLAPFvo7RpH2BhZlaD5I08GbC6\neUd54nRtRHQ7Kbgnisvh/145sYu2f0e2/6Cy3b6vun+fsl3VNu4aETuP0c5uXk4xgv9mj+0a8ZMo\nphC+Fa1TYRcALytPKJ8CvAm4rMsxux17PnBbefI2EMUl/NXATWO0/0+BDwDTy7I/B64r098APjlG\n3W7upPid7EBxpeSiiPi99TyGxltm+vI15gu4myI4fwD4J+Aw4ApgEpAUl0CDcgRRqfcS4K5y+0Lg\no5V9e5V1n12mvwB8ZJSfvy+wpJK+CvhAJf0O4Hs99qXbsart2htYAwxQnAD8DHh+2zF2A4aA7Sp5\n/wR8odz+EHBRuf0KYLDb77O9bJneleIDd5tK3vHAlaP05zbgVZX07wGPl+/NHuXveM/K/m55fw9c\nUklvBdwPvKLS3j9bj7+XNwNXt+WdNfL76VJ+P4qTqUnAERQj6ZeV+54K7F/u25Ui+FxeqftTikvc\nU8vjPAzcXu77/fJv8hUUJ4R/DwwD7yv33wkcVjnW5PJ3s0db+y4CPjRGfy9o79tY7Sr3H1S2aUeK\nk89fApPKfdtTnHwlxUnf9cBOlTYupJhemExxIrim7XdyErCirPsocOQYbf8C8LlK+q+A2yrp51Fc\nKaj+7a6gmFoYeT0K/HSMn3EDcHSvfz++6nk5stf6+DLFiPqttF3CpxgRPgW4trwUvBT4XpkPxajz\nvkr5e0b7IRHxlIg4LyLuiYhHgJ8AO7Zdyv5NZftRiqCwocdqb9dkilHNl4HLgYujmHo4OyIml315\nOFtHrvdQXMnYWM8sf/4Dld/jeRRXMEYr/5+VsrdRnIjsWilzX5d61bxnUHk/MnO43D9jlPLrsoIi\nYFVtzyiXwzPzusx8KDPXZualFPPsf1zuW5GZ88t9DwKnAa+JiJHjv4nisvd9wGfLuoNl3V9RXJ35\nv8ADFO/prSP7u7RzZHvMy/ZVUUxT/QnlJfyKUdtVtu0nmbkmM5dSzNHPorjSQll+KsUJ0LbAtyhH\n9pn5OMWVnCMp/g+8B7hk5NhRLJ47mydPcA4GPh8R+47RjQcr26u6pNv/b70uM3cceVGcbFd/J2+J\nYvHuyN/kH1D87rUZGezVs8y8h2LR0REUH0BViyk+GPapfBDskMXiJSg+bHerlN99jB/1HuA5wIGZ\nuT3FKAi6X8Jcl16O1d6ux4HFWcy3/kNm7g28FPgjikv+i4CdImK7tnr3d/n5KylOgoofWpxkVBeq\ntT928j6Kkf30yu9x+8wc7RL4fcDh1Q/fzJyamdW2dHu0ZTVvEcVJw0gbg+J3sq5jjOYWYM+2388L\nyvxeJKO/1yPtCCj+JjPzjzJzl8w8kCJAXvNE4cxvZOYfZObOwBkU/ZxXaecLnjw0LwAezMyHemwn\nFCclD1NcIXqyketo1yj9GunzCyiuFDycmasprhAcMDJ9kZk3ZebBmblzZh5KsbBw5Nj7Aj8pT5CG\nM3Me8AvGnjYbNxHxTOBzFCdlO5cnA79kw/7vahwZ7LW+TgL+MDNXVjPL0eDngH+JcuVvFF8/Gpkn\nvgR4a0TsXc5DnjHGz9iO4sRhabnYbKyy69LLsU6otOtM4BuZORQRr4yI55UB+hGKk4ChzLyP4vL+\nP0WxiO/5FL+Xr3Q59q+BqRFxZHlV4APA1pX9DwJ7RMRWAJn5APB94J8jYvtyAd6zIuLgjiMXzgXO\nKj9kiYhdIuLo9fj9QPHeHBkRryrb+B6KE46fredxKPvwa4pLt2eUv59jKFaWf7Nb+Yg4NiKeWvb1\nNcAJwNxy34ER8Zxy384U6ymuysxl5f7nRsR2ETElIk6guKz9ycqxX1jOXe9CcYXkO+WIH4qrUyeV\n7/00ivfmC5W6kyNiKsXn5KSyL+0LJU8EvpSZLSdDY7UrIvaJiH3Ldj0V+GeKE6vbyurzgLdEsVBz\nMsXIeVGWK/0j4vllW54SEX9DMXXzhUrdl4+M5KNYXPdyxp6zH0/bUpy4/K78+W+jc62CNgODvdZL\nZt6ZmfNH2f1eiq8rXV1eMv8BxaiaLFY1f4pi9feC8t/RfIria32LKRYGfm8jmtzLsb5M8WH5G4rL\np+8s859OMUf8CMUH8Y95clHi8RTz34uA/6RYeX5F+4HLoPQO4PMUH+grqVzOBb5e/vtQRFxXbr+F\n4hLsrRQL7r5B8YHezacpAuP3I2J52ccDRynbVWbeThFgP0Pxe3ot8NrMXDNanSi+aXHuGIc9jmKu\nfQnwUeDYzBwJAG+KiOoo/10Uv5ulwMeBt2fmVeW+PSnes+UUI8TVVL4CSrHocGH5c06hmIP/XWX/\np8vj3l7++8S3JrJYIHk2cCXFNMY9tJ4Mfo7iRPF4iq8PruLJVe8jX3P7QzqntNbVrl2Br1H8XS2k\n+Dv6o/ISPRSLTR8D7qAImkcA1RXtb6a4UvZbiq8Nvrq8AkBm/phiHcg3yr+HbwL/mJnf79LGcZeZ\nt1KcvPyc4kT2eRTfFtBmFm0npJIkqWEc2UuS1HAGe0mSGs5gL0lSw9Ua7CPisIi4PYrbiZ4+Rrlj\no7iF5/6VvPeV9W6vrOiWJEnrab0egrE+yq+onENxe9RBYF5EzC1Xa1bLbUex+vkXlby9KVbz7kNx\nw48fRMRe2flwDEmStA61BXvgAGBBZi4EiIiLgaMpvk5U9WGKr79U721+NHBx+XWSuyJiQXm8n4/2\nw6ZP3zn32H2s+7SMbu3QEJMGuj5npNH6sd/92Gew3/3GfjdUDndkXXvDTYszc6wnSgL1BvsZtN5i\nc5C27/+WN3zYLTO/W94colr36ra6Y96KdI/dd2f+T6/aoIYuXrKU6dN23KC6E1k/9rsf+wz2u9/Y\n7wboEth5vPPWFzHt6aPeeryqzmDf7faIT3ypv7xj2L9QeaRkr3UrxziZ4nnOzJwxg8VLlm5QQ5ct\nX7FB9Sa6fux3P/YZ7He/sd8N0C3Yr127wYerM9gP0nrP8ZkUdxsbsR3FbRSvKm7FzdOBuRFxVA91\nAcjM84HzAfbfb05uzBldY84G11M/9rsf+wz2u9/Y7wmux5F9r+pcjT8PmB0Rs6J4LvZxlPe7huI2\nopk5PTP3yMw9KC7bH1XeinUucFxEbB0Rsyiefz7WQyQkSdIoahvZZ+baiDiN4hGhA8CFmXlLRJwJ\nzM/MuWPUvSUiLqFYzLcWONWV+JKkfpFLf9uRFzuO9qTrdavzMj7ls6kvbcv74ChlX9GWPgs4q7bG\nSZLUJ7yDniRJDWewlySp4Wq9jC9JktYtH7q/JR3b79xZKDZ8fO7IXpKkhjPYS5LUcAZ7SZIazjl7\nSZI2obz/1x15sXPb418mTx3Xn+nIXpKkhjPYS5LUcAZ7SZIazmAvSVLDuUBPkqQa5dLftGbssEtn\noanb1toGR/aSJDWcwV6SpIYz2EuS1HDO2UuSNI5y6YOtGSuWtSRj5nM2YWsKjuwlSWo4g70kSQ1n\nsJckqeEM9pIkNZwL9CRJ2kC5aEFH3vBv7mpJD+z36k3VnFE5spckqeEM9pIkNZzBXpKkhnPOXpKk\nHuXiwZb08OAdHWUGDjh8UzWnZ47sJUlqOIO9JEkNZ7CXJKnhnLOXJKmLjgfaAMPX/qglPXDoWzZV\nczaKI3tJkhrOYC9JUsMZ7CVJajiDvSRJDecCPUmSgHxkcUt6+LKLOsoMHPfXm6o548qRvSRJDWew\nlySp4Qz2kiQ1nHP2kqS+k0t/05E39M9/15KedOYFnRUj6mpSrRzZS5LUcAZ7SZIazmAvSVLDOWcv\nSWq+Vctbko+//7SOIlP+pe179RN0fr4bR/aSJDWcwV6SpIYz2EuS1HAGe0mSGs4FepKkZlm1oiNr\n7ZmntqSnfPqrnfUmTamrRZudI3tJkhrOYC9JUsMZ7CVJajjn7CVJE9uax4p/H18Dax5j1Wlv7Ciy\nzbmXtGY0eH6+G0f2kiQ1XK3BPiIOi4jbI2JBRJzeZf8pEXFzRNwQET+NiL3L/D0iYlWZf0NEnFtn\nOyVJarLaLuNHxABwDvBqYBCYFxFzM/PWSrGvZua5ZfmjgE8Ch5X77szMfetqnyRJ/aLOOfsDgAWZ\nuRAgIi4GjgaeCPaZ+Uil/LZA1tgeSdIElyuWdOQ98obXAbByxjOZfP897PCN/+6sOHlq3U3botV5\nGX8GcF8lPVjmtYiIUyPiTuBs4J2VXbMi4vqI+HFEvLzGdkqS1Gh1juy7PRuwY+SemecA50TEG4EP\nACcCDwC7Z+ZDEfFC4NsRsU/blQAi4mTgZICZM2aweMnSDWrosuWdd1vqB/3Y737sM9jvftPofne5\nO97KGc8EYMX0XQF4fNnyjjI8trbWZm3p6gz2g8BulfRMYNEY5S8GPguQmauB1eX2teXIfy9gfrVC\nZp4PnA+w/35zcvq0HTe4sRtTdyLrx373Y5/BfvebpvY7J3fO9k6+/54ntne8/x522GG7zorbPLXO\nZm3x6ryMPw+YHRGzImIKcBwwt1ogImZXkkcCd5T5u5QL/IiIPYHZwMIa2ypJUmPVNrLPzLURcRpw\nOTAAXJiZt0TEmcD8zJwLnBYRhwCPA0soLuEDHAScGRFrgSHglMx8uK62SpK2TLmydXr2t0ce0VHm\nad8pxpGPP7KS7bfftu9H8d3Uege9zLwUuLQt74OV7XeNUu+bwDfrbJskSf3CO+hJktRwBntJkhrO\nB+FIkrYI3W6Y8+hfvKklvevlP+isOHXb4t+hycT2zfwWwsZyZC9JUsMZ7CVJajiDvSRJDeecvSRp\n08i2u9+tfrQlueCgQzqqPPuq77VmjMzPa704spckqeEM9pIkNZzBXpKkhjPYS5LUcC7QkyRtErns\nty3p+484qiU9+5qfd1aaNKXOJvUNR/aSJDWcwV6SpIYz2EuS1HDO2UuSxl0uHuzI+/kBh7WkX3rr\nL1oLOD9fG0f2kiQ1nMFekqSGM9hLktRwztlLkjZaPnR/S3regYd3lHnpLVe3ZvhQm03Gkb0kSQ1n\nsJckqeEM9pIkNZzBXpKkhnOBniRpveSS33Tk/fxFh7akX3r7/M6Kk6fW1SStgyN7SZIazmAvSVLD\nGewlSWo45+wlSWPKB+9qSf/wgCM7yhyy4LrWDOfntyiO7CVJajiDvSRJDWewlySp4ZyzlyQ9IRcP\nduTNf+lRLelD7riuo4xz9Fs2R/aSJDWcwV6SpIYz2EuS1HAGe0mSGs4FepLUx/Kh+1vS351zSEeZ\n1951U2vGpCl1Nkk1cGQvSVLDGewlSWo4g70kSQ3nnL0k9YkcvL0j79/2O6IlfergLZ0VnaOf8BzZ\nS5LUcAZ7SZIazmAvSVLDOWcvSQ2VDy9qSX/rxUd1lOmYo5/iA22ayJG9JEkNZ7CXJKnhDPaSJDWc\nwV6SpIZzgZ4kNUDLDXMeXU2ufJAL9z+ypcxJ9/6ys6IL8vpCrSP7iDgsIm6PiAURcXqX/adExM0R\ncUNE/DQi9q7se19Z7/aIOLTOdkqS1GS1BfuIGADOAQ4H9gaOrwbz0lcz83mZuS9wNvDJsu7ewHHA\nPsBhwL+Vx5MkSeupzpH9AcCCzFyYmWuAi4GjqwUy85FKclsgy+2jgYszc3Vm3gUsKI8nSZLWU51z\n9jOA+yrpQeDA9kIRcSrwbmAK8IeVule31Z1RTzMlaeIZvvfWlvS3X3rMk4m9ZsOv7+CkB37dWilc\nk92v6gz20SUvOzIyzwHOiYg3Ah8ATuy1bkScDJwMMHPGDBYvWbpBDV22fMUG1Zvo+rHf/dhnsN9N\nlI+uac3Ya/aT27vNBOj8TGx4sG/y+72x6gz2g8BulfRMYNEoZaG4zP/Z9ambmecD5wPsv9+cnD5t\nxw1u7MbUncj6sd/92Gew300zvLztsbO/vqMj3dH3hgd7aO77vbHqfOfnAbMjYlZETKFYcDe3WiAi\nKqeiHAmM/LXOBY6LiK0jYhYwG7imxrZKktRYtY3sM3NtRJwGXA4MABdm5i0RcSYwPzPnAqdFxCHA\n48ASikv4lOUuAW4F1gKnZuZQXW2VJKnJar2pTmZeClzalvfByva7xqh7FnBWfa2TpIkhH7yrI+/0\nPzi8JX324idvqrN46SNM33H7vrhsr974lyBJUsMZ7CVJajiDvSRJDeeDcCRpC9N+w5wbX3VcR5mz\nlyxozRiY/OT25Md8wI1aOLKXJKnhDPaSJDWcwV6SpIZzzl6SNrPhO69vSX/15X/akj7h3ls6K1Xn\n6KV1cGQvSVLDGewlSWo4g70kSQ1nsJckqeFcoCdJm9DwHfM78r568PEt6RMW3tBaYFLbs+ul9eTI\nXpKkhjPYS5LUcAZ7SZIazjl7SapRLmp9YM2Vh7ylo0zHTXOco9c4c2QvSVLDGewlSWo4g70kSQ3n\nnL0kjZPh267uyPvMQW9sSb9r0a2dFZ2jV80c2UuS1HAGe0mSGs5gL0lSwxnsJUlqOBfoSdIGGl5w\nXUv6glec0FHmXYNtN8yZPLXOJkldObKXJKnhDPaSJDWcwV6SpIZzzl6SejB8100dee94wdEt6XMf\n+lVnxa23qatJUs8c2UuS1HAGe0mSGs5gL0lSwzlnL0ld5G/vbkn/y4te11Hmsw9c35oxddsaWyRt\nOEf2kiQ1nMFekqSGM9hLktRwBntJkhrOBXqS+t7wvbd25F358mNb0u9Z1FmGKT7URhODI3tJkhrO\nYC9JUsMZ7CVJajjn7CX1nfzdvS3pb7/0mI4yf3zXja0Zk52f18TlyF6SpIYz2EuS1HAGe0mSGs45\ne0mNlosHO/KuOfCIlvQfL7ius6Jz9GqQdY7sI2KnTdEQSZJUj14u4/8iIr4eEUdERNTeIkmSNK56\nCfZ7AecDbwYWRMQ/RsRe9TZLkiSNl3UG+yxckZnHA38OnAhcExE/joiX1N5CSZK0Uda5QC8idgZO\noBjZPwj8FTAX2Bf4OjCrzgZK0voYvvGqJ7ZzDXz5qLd3lHnLfbe0ZkyaUnOrpM2rl8v4Pwe2B16X\nmUdm5rcyc21mzgfOHatiRBwWEbdHxIKIOL3L/ndHxK0RcVNE/DAinlnZNxQRN5SvuevbMUmSVOjl\nq3fPyczstiMzPzZapYgYAM4BXg0MAvMiYm5mVp8TeT2wf2Y+GhF/CZwNvKHctyoz9+2lE5IkaXS9\nBPv/6rIIfxkwHzgvMx8bpd4BwILMXAgQERcDRwNPBPvMvLJS/mqK6QJJkjSOegn2dwG7AP9Rpt9A\nMXe/F/A5irn8bmYA91XSg8CBY/yck4DLKumpETEfWAt8NDO/3UNbJfWRbjfMufHYdz6x/dizn8Vb\nBm/tKMPA5DqbJW1xegn2czLzoEr6OxHxk8w8KCJuGbUWdPtOftfpgIg4AdgfOLiSvXtmLoqIPYEf\nRcTNmXlnW72TgZMBZs6YweIlS3voTqdly1dsUL2Jrh/73Y99hgb3e8WqjqzHnv2sJ7bXzHgGi5cs\n66w30Oybhzb2/V6Hfu13L3r5i98lInbPzHsBImJ3YHq5b80Y9QaB3SrpmcCi9kIRcQjwfuDgzFw9\nkp+Zi8p/F0bEVcAcoCXYZ+b5FPcAYP/95uT0aTv20J3uNqbuRNaP/e7HPkMz+51DnR/u9y1o+Zhg\n+rQdOiv2wci+ie93L/q13+vSS7B/D/DTiLiTYrQ+C3hHRGwLfHGMevOA2RExC7gfOA54Y7VARMwB\nzgMOy8zfVvKnAY9m5uqImA68jGLxniRJWk9jBvuI2ApYCswGfp8i2P+qsijvU6PVzcy1EXEacDkw\nAFyYmbdExJnA/MycC3wceCrw9XIR4L2ZeRTwXOC8iBim+HrgR9tW8UvqQ/nbu1vS333hYR1lXvvr\n+U9sL162vC9G8dK6jBnsM3M4Iv45M18C3Li+B8/MS4FL2/I+WNk+ZJR6PwOet74/T5Ikderlpjrf\nj4jX+xAcSZImpl7m7N8NbAsMRcQqikv5mZnb19oySZI0LtYZ7DNzu03REEmSVI9eHoQTwJuAWZn5\n4YjYDfi9zLym9tZJ6lvti/EArnjR4S3p1955Q2fFKVOf3H5s7Ti3SpqYepmz/zfgJTz5tbkVFPe8\nlyRJE0Avc/YHZuZ+EXE9QGYuiQifBylJ0gTRy8j+8fIJdgkQEbsAw7W2SpIkjZteRvb/Cvwn8LSI\nOAs4FvhAra2S1Hfyoftb0jcffHRHmVff+vPWjOr8vKRR9bIa/ysRcS3wKoqv3b0uM2+rvWWSJGlc\n9PropzuAR0bKVx+MI0mStmy9fPXur4AzKJ5hP0R5Ux3g+fU2TZIkjYdeRvbvAp6TmQ/V3RhJ/SN/\n13pxcMGhx7Skn3/jLzorOUcvbZBeVuPfByyruyGSJKkevYzsFwJXRcR/A6tHMjPzk7W1SpIkjZte\ngv295WtK+ZIkSRNIL1+9+weAiNg2M1fW3yRJkjSeelmN/xLgAuCpwO4R8QLgLzLzHXU3TlIz5OLB\njrxH3vymlvTsX/xPa4HJLsaTxksvC/Q+BRwKPASQmTcCB9XZKEmSNH56CfZk5n1tWUM1tEWSJNWg\nlwV690XES4Esn3b3TsDb5UqSNEH0EuxPAT4NzAAGge8Dp9bZKEkTWy5vvQfX3Ucc01Fm1k+uas1w\njl6qTS+r8RcDb1pXOUmStGXqac5ekiRNXAZ7SZIazmAvSVLDjTpnHxHvHqui98aXBJAP3d+Rd8OL\nD29Jz/lllyfYbb1NXU2S1GasBXrblf8+B3gRMLdMvxb4SZ2NkiRJ42fUYF+5J/73gf0yc3mZ/hDw\n9U3SOkmStNF6mbPfHVhTSa8B9qilNZIkadz1clOdLwPXRMR/AgkcA3yp1lZJ2mK1z9E/9p6/6Cgz\n56aftWY4Py9tVr3cVOesiLgMeHmZ9bbMvL7eZkmSpPHS61fvngI8kpmfBgYjYlaNbZIkSeNoncE+\nIs4A3gu8r8yaDFxUZ6MkSdL46WXO/hhgDnAdQGYuiojtxq4iqQmG7721I2/J297ekt75sh92Vpzi\nQ22kLUkvl/HXZGZSLM4jIratt0mSJGk89RLsL4mI84AdI+LtwA+Az9fbLEmSNF56WY3/iYh4NfAI\nxd30PpiZV9TeMkmSNC7WGewj4mOZ+V7gii55kiRpC9fLAr1XU6zGrzq8S56kCS4fWdySXvuJD3WU\n6ViQ52I8aYs31lPv/hJ4B/CsiLipsms74Gfda0mSpC3NWCP7rwKXAf8EnF7JX56ZD9faKkmSNG5G\nXY2fmcsy827g08DDmXlPZt4DPB4RB26qBkqSpI3Ty5z9Z4H9KumVXfIkbUkyO9Ntebn0wY5qa8/8\n3y3pKR+7sPPYztFLE04v37OP8qY6AGTmML2dJEiSpC1AL8F+YUS8MyIml693AQvrbpgkSRofvQT7\nU4CXAvcDg8CBwMl1NkqSJI2fXu6g91vguE3QFknjJjvSuey3LTlD553VUWvyx7/UmjFpyji3S9Lm\nMNb37P9PZp4dEZ+h85ODzHxeQ9WzAAAVuElEQVRnrS2TJEnjYqyR/W3lv/M3RUMkSVI9Rg32mfmd\n8t8vbrrmSJKk8TbWZfzv0OXy/YjMPKqWFkmSpHE11mX8T5T//jHwdOCiMn08cHcvB4+IwyjuwDcA\nfD4zP9q2/93AnwNrgd8Bf1bepY+IOBH4QFn0I15hkMbQfsOcFUta969aydBnz2jJGvibj9LBBXlS\nI411Gf/HABHx4cw8qLLrOxHxk3UdOCIGgHMonpo3CMyLiLmZeWul2PXA/pn5aPngnbOBN0TETsAZ\nwP4UVxeuLeu2fYJJkqR16eV79rtExJ4jiYiYBezSQ70DgAWZuTAz1wAXA0dXC2TmlZn5aJm8GphZ\nbh8KXJGZD5cB/grgsB5+piRJatPLbW//GrgqIkbumrcH8Bc91JsB3FdJj9yQZzQnUTxlb7S6M3r4\nmZIkqU0vN9X5XkTMBn6/zPpVZq7u4djR7XBdC0acQHHJ/uD1qRsRJ1PezW/mjBksXrK0h2Z1WrZ8\nxQbVm+j6sd+N7fOq1n4Nfe2zLelHnrIj/FXbHH1u3XmcDfw/tKVq7Pu9DvZb7dYZ7CPiKcC7gWdm\n5tsjYnZEPCczv7uOqoPAbpX0TGBRl+MfArwfOLhyEjEIvKKt7lXtdTPzfOB8gP33m5PTp+24ru6M\namPqTmT92O8m9jnb1tUNPfK7jjLTd9y+NWPrbWps0Zajie93L+y3qnqZs/93YA3wkjI9CHykh3rz\ngNkRMSsiplDccndutUBEzAHOA44qb8s74nLgNRExLSKmAa8p8yRJ0nrqJdg/KzPPBh4HyMxVdL/M\n3iIz1wKnUQTp24BLMvOWiDgzIka+o/9x4KnA1yPihoiYW9Z9GPgwxQnDPODMMk+SJK2nXhborYmI\nbSjnzCPiWUAvc/Zk5qXApW15H6xsHzJG3QuBC3v5OVJfWdU5Lzn06fe3pCf93b+2pAeWPtI3l+0l\ndeol2J8BfA/YLSK+ArwMeGudjZIkSeNnzGAfEQH8iuIuei+muHz/rsxcvAnaJkmSxsGYwT4zMyK+\nnZkvBP57E7VJkiSNo14W6F0dES+qvSWSJKkWvczZvxI4JSLuBlZSXMrPzHx+nQ2T+lYOtyYfXd6S\nHvrkezuqTHrfv7ZltH3xftJj49I0SRNTL8H+8NpbIUmSajPW8+ynAqcAzwZuBi4ovzsvSZImkLHm\n7L9Icb/6mylG9/+8SVokSZLG1ViX8ffOzOcBRMQFwDWbpklSg2Xb85yi82aUuWJJS3r4e19pSU/6\nm090Hrd9jl6SKsYa2T8+suHle0mSJq6xRvYviIhHyu0AtinTI6vxtx+9qiRJ2lKMGuwzc2BTNkSS\nJNWjl6/eSRovbXP0uXJpR5Hh736hJT3w+lNbC0yZOt6tktRwvdxBT5IkTWAGe0mSGs5gL0lSwxns\nJUlqOBfoSZvSqtaH2gx/8eMdRbZ687tbM1yQJ2kjObKXJKnhDPaSJDWcwV6SpIZzzl6qSfsDbQCG\nL/tyS3rgz/++s6Jz9JLGmSN7SZIazmAvSVLDGewlSWo45+yl8bJqRUty+PorO4oMHHtaa0Z4vi2p\nfn7SSJLUcAZ7SZIazmAvSVLDGewlSWo4F+hJGyBXLu3IG77qmy3pgcPe0lnRBXmSNgM/eSRJajiD\nvSRJDWewlySp4Zyzl9plduY91nbDnPlXdBQZeNVxbRmTx7NVkrTBHNlLktRwBntJkhrOYC9JUsM5\nZy+1a5ufBxi69gct6YH/9brOes7RS9pCObKXJKnhDPaSJDWcwV6SpIYz2EuS1HAu0FOztd8gJ4eL\nV9Wa1S3J4Ttv6DjMwMuObs3wgTaSJhA/sSRJajiDvSRJDWewlySp4ZyzV7NFtKbXrIbHHm3JGv71\n/Jb0Vnu/uMtxPC+WNHH5CSZJUsMZ7CVJajiDvSRJDeecvRotVy5tTf/uPoZXtP7Zb/UH/6u10oD/\nLSQ1S60j+4g4LCJuj4gFEXF6l/0HRcR1EbE2Io5t2zcUETeUr7l1tlOSpCarbQgTEQPAOcCrgUFg\nXkTMzcxbK8XuBd4K/E2XQ6zKzH3rap8kSf2izuuVBwALMnMhQERcDBwNPBHsM/Puct9wtwNIkqSN\nV2ewnwHcV0kPAgeuR/2pETEfWAt8NDO/3V4gIk4GTgaYOWMGi5csbS/Sk2XLV2xQvYmuL/q9qrWP\ny9YMdRSJJctaMwYG6mzRZtEX73UX9ru/9Gu/e1FnsI8uedklbzS7Z+aiiNgT+FFE3JyZd7YcLPN8\n4HyA/febk9On7bjBjd2YuhNZo/r92MqOrKGFt7ZmzNq/s899csOcRr3X68F+95d+7fe61PkpNwjs\nVknPBBb1WjkzF5X/LgSuAuaMZ+MkSeoXdQb7ecDsiJgVEVOA44CeVtVHxLSI2Lrcng68jMpcvyRJ\n6l1twT4z1wKnAZcDtwGXZOYtEXFmRBwFEBEviohB4E+A8yLilrL6c4H5EXEjcCXFnL3BXpKkDVDr\n3UMy81Lg0ra8D1a251Fc3m+v9zPgeXW2TRNMdlnuMfR4S3L419d2FBnY95WtGY+s7Js5ekka4aee\nJEkNZ7CXJKnhDPaSJDWcwV6SpIbz8V6aGFY/2pGVD93fkt5qn5d01huYXFeLJGnCcGQvSVLDGewl\nSWo4g70kSQ3nnL22TG0PtcmVnU80jF12b81wfl6SunJkL0lSwxnsJUlqOIO9JEkN55y9tgztc/SL\nB1vS8XvP6qwz4J+vJPXCkb0kSQ1nsJckqeEM9pIkNZzBXpKkhnOFk+qX2Zpeu7qzyKPLWtLx9Fmt\nBVyMJ0kbzJG9JEkNZ7CXJKnhDPaSJDWcE6EaX+3z80VmS2r4gYUdJbaasVdrhnP0kjRuHNlLktRw\nBntJkhrOYC9JUsM5Mapx1jlnn8sfbkl3zM+Dc/SSVCNH9pIkNZzBXpKkhjPYS5LUcAZ7SZIazlVR\n6l0PN8zhsZUdJWLrbVszXIwnSZuUI3tJkhrOYC9JUsMZ7CVJajgnTzW69jn6occ7i6xY0pKOHZ7W\neZyI8WyVJGk9ObKXJKnhDPaSJDWcwV6SpIZzzl6FHF53kdWPduTFdju3ZTg/L0lbGkf2kiQ1nMFe\nkqSGM9hLktRwBntJkhrOBXoqDHdZoLfmsZZkPGX7zjLh+aIkben8pJYkqeEM9pIkNZzBXpKkhnPO\nvh+0P9AGYHgIhoZgaG2RbpufB2DqU1rTzs9L0oTkp7ckSQ1nsJckqeEM9pIkNVytwT4iDouI2yNi\nQUSc3mX/QRFxXUSsjYhj2/adGBF3lK8T62xn4w0Pdbxy1XJY8yi5anmxPWXrzlds1fqSJE1ItX2C\nR8QAcA5wOLA3cHxE7N1W7F7grcBX2+ruBJwBHAgcAJwREdPqaqskSU1W53DtAGBBZi7MzDXAxcDR\n1QKZeXdm3gS0377tUOCKzHw4M5cAVwCH1dhWSZIaq85gPwO4r5IeLPPqritJkirq/J59dMnr8oXv\nDa8bEScDJwPMnDGDxUuW9t66imXLV2xQvQljaKgzb82jLFu56sn0Y13KDDTvNgyNf69HYb/7i/1W\nuzo/zQeB3SrpmcCi9aj7ira6V7UXyszzgfMB9t9vTk6ftuOGtBOAjam7xRl6vDU93CWQT9oJlix9\nst/R7fyqmRr1Xq8H+91f7Leq6ryMPw+YHRGzImIKcBwwt8e6lwOviYhp5cK815R5kiRpPdUW7DNz\nLXAaRZC+DbgkM2+JiDMj4iiAiHhRRAwCfwKcFxG3lHUfBj5MccIwDzizzJMkSeup1knZzLwUuLQt\n74OV7XkUl+i71b0QuLDO9kmS1A+atwKr6bo91KbdmtWt6fYH2kAxRz/ykiQ1mrdFkySp4Qz2kiQ1\nnMFekqSGc85+omn7znyuWt5RJLbZri3DczpJ6mdGAUmSGs5gL0lSwxnsJUlqOIO9JEkN5wK9LV37\nTXSG1rYkY9sdOuu4IE+SVGFUkCSp4Qz2kiQ1nMFekqSGc85+S5LDnXnDbXkDbW+Z8/OSpHUwUkiS\n1HAGe0mSGs5gL0lSwzlnvyVpn5+Hzjl6SZLWkyN7SZIazmAvSVLDGewlSWo4g70kSQ3n6q9Npf2B\nNpIkbSKO7CVJajiDvSRJDWewlySp4ZyzHw+9zMdHdOYNrW1NewMdSVINHNlLktRwBntJkhrOYC9J\nUsMZ7CVJajhXhNWmbdHe0FBnERfkSZI2AUf2kiQ1nMFekqSGM9hLktRwThpviBxed5loO48a8LxK\nkrR5GIEkSWo4g70kSQ1nsJckqeGcs1+Xbg+5Ge5hzt45eknSFsKIJElSwxnsJUlqOIO9JEkNZ7CX\nJKnh+nuB3sjiu8wntyPWXW+rgdZ0L3UkSdpMHNlLktRwBntJkhrOYC9JUsP195z9yFx7xJPb3W6i\nM1o9SZImAEf2kiQ1nMFekqSGM9hLktRwtQb7iDgsIm6PiAURcXqX/VtHxNfK/b+IiD3K/D0iYlVE\n3FC+zq2znWMamc+vviRJmkBqW6AXEQPAOcCrgUFgXkTMzcxbK8VOApZk5rMj4jjgY8Abyn13Zua+\ndbVPkqR+UefI/gBgQWYuzMw1wMXA0W1ljga+WG5/A3hVhENnSZLGU51fvZsB3FdJDwIHjlYmM9dG\nxDJg53LfrIi4HngE+EBm/r/2HxARJwMnA8ycMYPFS5ZuUEOXLV/xZKL9q3cNPvdo6Xef6Mc+g/3u\nN/Zb7eoM9t2iZPuX2Ecr8wCwe2Y+FBEvBL4dEftk5iMtBTPPB84H2H+/OTl92o4b3Ngn6vZRsIdK\nv/tIP/YZ7He/sd+qqvMy/iCwWyU9E1g0WpmImATsADycmasz8yGAzLwWuBPYq8a2PsnFeJKkhqkz\n2M8DZkfErIiYAhwHzG0rMxc4sdw+FvhRZmZE7FIu8CMi9gRmAwtrbKskSY1V22X8cg7+NOByYAC4\nMDNviYgzgfmZORe4APhyRCwAHqY4IQA4CDgzItYCQ8ApmflwXW2VJKnJar03fmZeClzalvfByvZj\nwJ90qfdN4Jt1tk2SpH7RrAfh5PCT2+HNASVJAm+XK0lS4xnsJUlqOIO9JEkNF9l+E5kJKiJ+B9yz\ngdWnA4vHsTkTRT/2ux/7DPa739jv/vHMzNxlXYUaE+w3RkTMz8z9N3c7NrV+7Hc/9hns9+Zux6Zm\nv9XOy/iSJDWcwV6SpIYz2BfO39wN2Ez6sd/92Gew3/3GfquFc/aSJDWcI3tJkhqu0cE+Ig6LiNsj\nYkFEnN5l/9YR8bVy/y8iYo8yf4+IWBURN5Svczd12zdGD/0+KCKui4i1EXFs274TI+KO8nVie90t\n2Ub2e6jyfrc/nXGL1kO/3x0Rt0bETRHxw4h4ZmVfk9/vsfrd5Pf7lIi4uezbTyNi78q+95X1bo+I\nQzdtyzfOhvZ7on+ej5vMbOSL4kl7dwJ7AlOAG4G928q8Azi33D4O+Fq5vQfwy83dhxr7vQfwfOBL\nwLGV/J0oHiW8EzCt3J62uftUd7/LfSs2dx9q7PcrgaeU239Z+Ttv+vvdtd998H5vX9k+Cvheub13\nWX5rYFZ5nIHN3adN0O8J+3k+nq8mj+wPABZk5sLMXANcDBzdVuZo4Ivl9jeAV0VEbMI21mGd/c7M\nuzPzJmC4re6hwBWZ+XBmLgGuAA7bFI0eBxvT74msl35fmZmPlsmrgZnldtPf79H6PZH10u9HKslt\ngZGFWUcDF2fm6sy8C1hQHm8i2Jh+i2Zfxp8B3FdJD5Z5Xctk5lpgGbBzuW9WRFwfET+OiJfX3dhx\n1Eu/66i7uW1s26dGxPyIuDoiXje+TavV+vb7JOCyDay7JdmYfkPD3++IODUi7gTOBt65PnW3UBvT\nb5i4n+fjplmPuG3VbYTefqY3WpkHgN0z86GIeCHw7YjYp+3McUvVS7/rqLu5bWzbd8/MRRGxJ/Cj\niLg5M+8cp7bVqed+R8QJwP7Awetbdwu0Mf2Ghr/fmXkOcE5EvBH4AHBir3W3UBvT74n8eT5umjyy\nHwR2q6RnAotGKxMRk4AdgIfLy1wPAWTmtRRzRXvV3uLx0Uu/66i7uW1U2zNzUfnvQuAqYM54Nq5G\nPfU7Ig4B3g8clZmr16fuFmpj+t3497viYmDkykXj3++KJ/o9wT/Px8/mXjRQ14viqsVCioUoIws6\n9mkrcyqtC/QuKbd3oVy4QrEg5H5gp83dp/Hqd6XsF+hcoHcXxWKtaeV2P/R7GrB1uT0duIO2xT9b\n6qvHv/M5FB9ws9vyG/1+j9Hvpr/fsyvbrwXml9v70LpAbyETZ4HexvR7wn6ej+vvcHM3oOY/kCOA\nX5f/4d9f5p1JcZYPMBX4OsVClWuAPcv81wO3lH9Q1wGv3dx9Ged+v4jiTHkl8BBwS6Xun5W/jwXA\n2zZ3XzZFv4GXAjeX7/fNwEmbuy/j3O8fAA8CN5SvuX3yfnftdx+8358uP79uAK6sBkWKqxx3ArcD\nh2/uvmyKfk/0z/PxenkHPUmSGq7Jc/aSJAmDvSRJjWewlySp4Qz2kiQ1nMFekqSGa/Id9KS+EhE7\nAz8sk08HhoDflekDsrinuKQ+5FfvpAaKiA9RPNntE235QfH/fot7GFBETMriGRWSxpmX8aWGi4hn\nR8Qvy+d4XwfsFhFLK/uPi4jPl9u7RsS3yofEXBMRL+5yvEkR8cly/00R8edl/iHlc+O/VT53/EuV\nOi8qH0JybURcFhG7lvk/jYizIuInwGkRMTsiflEe+8Mj7YyI/4iIIyvH+1pEHFHTr0xqHIO91B/2\nBi7IzDkUtwsdzb8CZ2fm/sCfAp/vUuZk4LeZeQDFXQlPjYjdy337UdyGem/guRHx4ojYmuLuZq/P\nzBcCFwEfrhxv+8w8KDM/BXwG+ER57AcrZT4PvA0gIqaVP/fy3rsv9Tfn7KX+cGdmzuuh3CHAc4qr\n/QBMi4htMnNVpcxrKAL5cWV6B2B2uX11Zj4AEBE3AHsAj1Hcl/0H5XEHKG5bPOLiyvaBFLdFBfgq\n8JFy+0fAZ8p1CcdTPMdiqIf+SMJgL/WLlZXtYVofGTq1sh2sezFfAO/IzB+2ZBZPmFtdyRqi+IwJ\n4KbMHO054itHyX9CZmZEfAV4I/DW8l9JPfIyvtRnysV5S8r58a2AYyq7f0BxGR6AiNi3yyEuB95R\nPhaaiHhORGwzxo+8FZgREQeU5adExD6jlL2m0p7j2vb9O/C3wGOZefsYP09SG4O91J/eC3yP4qt6\n1UvqpwIvKxfe3Qq8vUvd8ygeC3tDRPwS+CxjXCXM4jnyxwKfjIgbgespLtd3807gvRFxDfA0YFnl\nOIsonnr27z31UNIT/OqdpC1GRGwLPFpetj8BOCYzX1/ZdzPwgsxcvjnbKU00ztlL2pK8CPhUOb2w\nhCdX4B8KfA74uIFeWn+O7CVJajjn7CVJajiDvSRJDWewlySp4Qz2kiQ1nMFekqSGM9hLktRw/x8a\n439NSsqg5AAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1cbc040a50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1,1, figsize=(8,8))\n",
    "counts, xedges, yedges = np.histogram2d(test_labels.flatten(), prediction.flatten(), bins=100)\n",
    "\n",
    "ax.pcolormesh(xedges, yedges, counts.T, cmap='Reds')\n",
    "\n",
    "ax.set_xlabel(\"True energy\")\n",
    "ax.set_ylabel(\"Predicted energy\")\n",
    "ax.grid(alpha=0.3)\n",
    "ax.set_title(\"Median absolute error: {0} mHa\".format(np.median(np.abs(prediction.flatten() - test_labels.flatten()))*1000.))\n",
    "fig.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}