WeightDistrubution-Hessian-FC-MNIST.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-20T20:13:59.673640Z",
     "start_time": "2019-01-20T20:13:57.997310Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialized TensorFlow\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "from experiment_train import *\n",
    "from experiment_datasets import *\n",
    "from helpers import *\n",
    "import matplotlib"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_H(scale = 1.0, epochs = 15, N = [100, 50]):\n",
    "    # scaling the architecture up/down\n",
    "    N = [int(t * scale) for t in N]\n",
    "\n",
    "    def get_p_arr(p):\n",
    "        \"\"\" p array with failure on the first layer \"\"\"\n",
    "        return [0, p]  + [0] * len(N)\n",
    "\n",
    "    # Lips. coeff\n",
    "    KLips = 1.\n",
    "\n",
    "    # activation function\n",
    "    activation = 'sigmoid'\n",
    "\n",
    "    # training the network\n",
    "    exp = MNISTExperiment(N = N, p_inference = get_p_arr(0), p_train = get_p_arr(0), KLips = KLips, epochs = epochs,\n",
    "                                      activation = activation, reg_type = None, reg_coeff = 0, do_print = 'plot',\n",
    "                                      name = 'experiment_weights', seed = None, batch_size = 100)\n",
    "    \n",
    "    # returning the weights in the middle\n",
    "    return exp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAD6CAYAAACvZ4z8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nO3deXxU5dn/8c+VfV9IwhogoCyyRSAsioo/KBZtC4qluFasy2MrbtWnD1ZbrctT21prba0WKyrqU0VbEeu+QLWCSEAWgYRdCOtk38h+/f44kzCEhAwwYZKZ6/16zWvOnDln5joavrlzn/vcR1QVY4wxgSvE3wUYY4xpXxb0xhgT4CzojTEmwFnQG2NMgLOgN8aYAGdBb4wxAc6roBeRqSKSKyJbRWRuC+/3FZGPRWSdiCwVkXSP934rIhtEZJOIPCEi4ssDMMYYc2xhbW0gIqHAk8AUIA9YKSKLVXWjx2aPAgtU9QURmQT8GrhaRM4GJgAj3Nv9B5gILG3t+1JTUzUjI+MEDsUYY4LXqlWr8lU1raX32gx6YCywVVW3A4jIK8B0wDPohwA/dS8vARa5lxWIAiIAAcKBA8f6soyMDLKzs70oyxhjTCMR+aa197zpuukF7PZ4nede52ktMMO9fAkQLyIpqrocJ/j3uR/vq+ombws3xhhz8nx1MvYuYKKIfIXTNbMHqBeR04EzgHScXw6TROTc5juLyI0iki0i2S6Xy0clGWOMAe+Cfg/Q2+N1untdE1Xdq6ozVHUkcI97XTFO6/4LVS1X1XLgXeCs5l+gqvNUNUtVs9LSWuxiMsYYc4K8CfqVwAAR6SciEcBlwGLPDUQkVUQaP+tuYL57eRdOSz9MRMJxWvvWdWOMMadQm0GvqnXAHOB9nJBeqKobROQBEZnm3ux8IFdENgPdgIfd618HtgHrcfrx16rqW749BGOMMcciHW2a4qysLLVRN8YYc3xEZJWqZrX0nl0Za4wxAc6bcfTGGGNOlirUVkJVKVSVNHsUO88xKZB1rc+/2oLeGGO8VV8HhwqhIv/IoK4uPRzWx3o01B3789PHWNAbY4xPqUJ1GVS4nPCucLXw8FhfWYhzwX8rwqIhKvHwIyYVupx25LojHknu5wSITIDwqHY5TAt6Y0xgaah3h/NBKD/QLMDdy+UHDy/XV7f8OVGJEJvmPFIHQt8J7tepThdLdNKRYR2ZAGERp/ZYvWRBb4zp+FThUJET0OUH3GF9wP046PE4AJX5oA1Hf0Zo5OGgjk2DbkMPL3uuj01zWuIdNLRPhAW9McZ/GhqccC7eBWX73C3tgx4B7hHkDbVH7x8aAXHdIK4rJPWGXqMOv258bgzvyHgI0lnSLeiNMe1H1Wl9F++Cop3Oc/EuKP7G/bz76K4TCXGCuTGsuw5xv/YM8G4Ql+Z0mwRpeB8PC3pjzIlTdU5QFn/jEd67oMhjue7QkfvEpEBSH+g2DAZd5Cwn9YWEnk6Ax3SBkFD/HE+AsqA3xrSuocFpkZfkQcku57mpVe5+1JQfuU9UkhPeaQNhwBQnxJP6HH5ExvnnWIKYBb0xway2Ckr3QMlud4jvPjLUS/Yc3bUSmeCEd3I/6DcRkpsFeVSif44lAKgq7XG3VQt6YwJV40iVkt0eAb77yFCvONhsJ4H47pDYG3qcCYO/64R3YrqzLjHdGVbYgRVW1JCzv5TN+8vYWVBJZHgICVHhJESHkxgdTkJUGAnR4e51YSRGhxMZ5vuuIlXlUG09BeU1FFXWUFBRQ6HHclGFe53H8qDu8Sz8r6Nmcj9pFvTGdEZ11VC23/3Y6zyXup/L9jmP0n1QW3HkfmFRhwN70NTDy43PCb06zbDC8uo6Nh8oY/P+MnIPlLH5QBm5+8vJLz/8F0hcZBi19Q1U17Uw3NJDZFiIO/wP/xJIjHZ+ETT+kvBcFx4aQnFlDYUVtRRWVB8R3EWVTqAXVNS0+r3hoUJyTARdYp3HkJ4JpMRGcHrX9unWsqA3piNpaHDGgTcGddm+I4O7MdgrC47eNzTCaY3H93ROdA644MgQT+rjnAjtZKNUquvq2Xawwglyj2DPKzp8kjc6PJSB3eOZNDiNgd3iGdQ9nkHd4kmLj0REqKqtp7SqltJDde7nWkoO1VJaVUfpodrD77mXiytr2FVY6WxzqJa6hmPP8hsXGUZybDhdYiPpGh/FoG4JpMS5g7wx0OPcy3ERxEeGtUsXTWss6I051apKIH8rFGyB/M1QsNVpjZfug/L9LcyHIs6wwvjukNgL0rMgvgck9HCeGx8xXdo1xFWVsuo6qmrrCRUhNEQICZHDy03PnFCI1Tco3xRUNLXMNx8oI2d/KTsLKql3B214qHBaWhyj+iRz+dg+Tqh3iyc9OZqQkNa/Myo8lKjwULrGn9hxH6qtb/olUXKolpq6BpJiwkmJjSQpJpyo8I49SsiC3pj20NAApXlOkOdv8Xje4oR5IwmFLv2cFne/cw+HtmeIx3WF0HCflVZT10DxoRpKD9VSXOk8Sg7VUnyolpLKmqblxvUlh5wWbmlVXVPgtiVEOCL8Q8X9S8G9rvn7IQL7SqqaujpEoG+XGAZ2i+ei4T0Y2C2ewd3jyUiNJTz01M6uLiLERIQRExFG98T2mYumvXkV9CIyFfgjEAr8TVUfafZ+X5zbB6YBhcBVqprnfq8P8Dec+84qcJGq7vTVARjjVzUVTou8McQbA71g65Hjx6MSnflSTp8MqQOc5ZQBkJxx0n3iVbX1uMqqOVhWxcHSag66lwvKa5wgP+Q8l7oDvLKmvtXPEoGEqHCSYsJJinb6pnt3iSHJfSIzKSacyPBQGhqU+galQZ3nelX3Og4vN61zllVptq3HssKUIZHuQE/g9K5xREd07FZyZ9Jm0ItIKPAkMAXIA1aKyGJV3eix2aPAAlV9QUQmAb8Grna/twB4WFU/FJE44NhnRYzpiKrLYN9aOLjJHeTuYC/Z7bGROEMNUwdC/4mQcrqznDrQmUflOLozVJXy6jontEud4HbCvJqDpVXuMHeWS6uOnvo2NERIiY0gOSaCxBbCOtG9PqnxdXQ4SdERxEeFHbMLxHRO3rToxwJbVXU7gIi8AkwHPIN+CPBT9/ISYJF72yFAmKp+CKCqza6sMKYDqq2CA1/DntWwd7XznL+ZpulpI+KcEO9zFqRe426hD3Cmo/Vymtmq2np2FVay3VXBzoIK9pdUHW6Vu8P9UO3RLe/IsBC6Jjgn/AZ0jWPCaSl0TYgiLT6SrvHO+q4JkXSJibDANk28CfpegGezJQ8Y12ybtcAMnO6dS4B4EUkBBgLFIvJPoB/wETBXVVv/29GYU6m+Dlw5hwN972o4sPHwBFqxXZ2JsobNgJ6joPswp9/ci9Z5Q4Oyt+QQ210V7Mh3HtvzK9juKmdP8SE8b9ccFxlG1/hI0uIjyUxPckLbHeiNy2nxUSREndrRGiYw+Opk7F3An0VkNvApsAeod3/+ucBIYBfwKjAbeNZzZxG5EbgRoE+fPj4qyZhmGhqgcLsT5nu/coJ939rDfemRidDzTDh7jhPqvUY548rbCNaiihq255cfGejulrrnOOq4yDD6pcYyqk8y3x+dTr/UWPqnxpGRGkN8lO9OthrTnDdBvwfnRGqjdPe6Jqq6F6dFj7sf/lJVLRaRPGCNR7fPImA8zYJeVecB8wCysrK8O61vzLGoOpf2e3a/7F0D1SXO+2HR0GMEjJ7tBHrPUdClP4QcPaKjuq6ewooa8stq2F1U2RTk2/PL2ZFfQXHl4elzw0KEPikx9E+NY+KgNPqlxjqBnhZLWlyktcaNX3gT9CuBASLSDyfgLwOu8NxARFKBQlVtAO7GGYHTuG+SiKSpqguYBGT7qnhjUHXmKs/ffORQxv3rD1/eHxLmTHU77BLoOYqGHiMpiT+dgkP15JfXUFBeQ8GWavLLt1JQXk1BeQ355c7Vjvnl1ZS1cLKze0IU/dNi+c7wHk1B3j81jvTkaMJO8fA/Y9rSZtCrap2IzAHexxleOV9VN4jIA0C2qi4Gzgd+LSKK03Vzs3vfehG5C/hYnKbMKuCZ9jkUE9Dq65z5zPM3Q36ux1DGzc4FSG4aHktlQn8OJo1nd4/BbAkbyMaG3hyoFPK3VVOwrobCijzqG3Yf9RUi0CUmgpS4CFJiIxnaM4HUuEhS4yJIiYukS2wE6cnRZKTEEhtpl6CYzkNUO1ZPSVZWlmZnW6M/aFWVHh662Bjkrs1O37rnHYbiukPqAKoST2OHpLPmUBpLCpL5974wqj1O9cdGhJISF9kU3k5oO8spcRGkxUU2vZ8cE0GojVQxnZSIrFLVrJbes2aJ8Z/8rbB9CbhyD3e7lO09/H5ImNNvnjoQBl+EpgwgL6w3X5amsHxvHau+KWJHjjNpV3ioMLxXItdM6MKoPslNrXG76MYYC3pzKqnCgQ2w6S3YtBgOui/FiExwxqH3n3j4AqPUgVTF92Hdvkqyvylk1c4iVi0roriyDCgjOSac0X2T+UFWb7IykhneK7HDzzdijL9Y0Jv2peqMetm42An3wu2AQN+zYepvYNCFzqyKIrjKqln1TSHZ24rI/sTFhr1bqa13uhb7p8VywZBuZPXtwuiMZPqnxtoIFmO8ZEFvfK+hAXavcIJ901vONAESivY7j6oxPyE/fQoFJFFUWcPezYdY9c1aVn1TxDcFlQBEhIWQmZ7Idef0J6tvMqP6JtMltnPMkW5MR2RBb05IbX0DRZXOZFlFFTUUlx8iIm8ZaXnvk+FaQlxtAbWEszZiJJ9EXcK7NSPZlRNF/UYFNhzxWalxEYzum8yV4/owum8XhvVKaJc7/hgTrCzoTatq6xtYl1fCih0FrP6miAOl1U3hXl5dRwS1TAj5mgtDvmRK6CqSpZxKjeTfeiafh1/DxrhxRMYmkRwbzlkxEVwUE05yTARJMREkx4STFBNB1/hI0pOjrRvGmHZkQW+aVNfVO8G+vYAVOwpZ9U1R05S2p3eNo1dSNGekhjG6dhWZZZ/Sv+g/RNSVUxceR3nfKRQM/h7RZ0xhakw8F1pwG9NhWNAHsaraetbsLmbF9kJW7Chg1TdFTXOzDO4ezw+yejOuXxfG9Qyjy95/w8Y3YetHUFsJ0V1g+MVwxnTC+k8kKSzSz0djjGmNBX0QOVRTz1e7ivhiRyErthfw1e5iauoaEIEzuidwxbg+jO+fwtiMLiTXuSD3HfjqbXjjP87FSnHd4Mwr4IzvQd9zINR+fIzpDOxfagCrrHEuKlqxvZAvthewNq+Y2nolRGBoz0R+OL4v4/unMCajC4nRYc649pwX4fO3ndkdwZl3/ayfwKDvQPqYFif9MsZ0bBb0AaSypo6VO4v4YnsBX2wvYH1eCXUNSmiIMKxXIj+a0I9x/buQldGFhKhwZ/6Y3Svg07ch921nLhlwAn3yfTD4u5A20K/HZIw5eRb0nVh1XT1rdhXz+bYClm/LZ81up8UeFiKMSE/khvP6M66fE+xxjZNw1VTCtvedbpncd+FQIYRGQP/zYcLtzgVM8d39eVjGGB+zoO9E6huUr/eU8Pm2fJZvK2DlzkKqap0+9uG9EvnROf04+7RUsvomHzm7YkU+bHwPct6GbUucG21EJcKAb8Pg7zg3rI6M99+BGWPalQV9B6aq5B4oY9nWApZtK2DFjoKmudEHdovjsjF9OPu0FMb1SyExptkdigq2Oa32nHdg9xegDZCQDqN+CIMvgr4TINTuamRMMLCg70BUlW8KKlm2rYBl7lZ7QUUNAH1TYvjuiB6cdVoqZ/VPIS2+heGMh4rhy2fg63+Aa5OzrttwOO+/nZZ79xFe3evUGBNYLOj9bH9JFcu25bNsWwHLtxWwp9i5f2m3hEjOG5jGWaelcPZpKaQnx7T+IYeKYPlfYMXTUF0KGefC6Edg0EWQ3PcUHYkxpqPyKuhFZCrwR5w7TP1NVR9p9n5fnNsHpgGFwFWqmufxfgKwEVikqnN8VHunVVffwPPLdvJ/X+5iu8uZTz0pJpyz+qdw0/mncfZpKd7NzlhZCF/8BVb81Qn4M74HE/8Hug8/BUdhjOks2gx6EQkFngSmAHnAShFZrKobPTZ7FFigqi+IyCTg18DVHu8/iHOLwaC3YW8Jc/+xnvV7ShjfvwtXjO3DWaelcEb3BEK8vbtRZSEsf9IJ+JoyGDIdzvsZdB/WvsUbYzolb1r0Y4GtqrodQEReAabjtNAbDQF+6l5eAixqfENERgPdgPeAFm9zFQyqaut54uMt/PXT7STHhPPkFaO4aHj345vMq6IAlv8ZvpwHNRVOwE/8GXQb2n6FG2M6PW+CvhfgeSflPGBcs23WAjNwuncuAeJFJAUoAn4PXAV866Sr7aRWbC/g7n+uZ3t+Bd8fnc693zmDpJjjmF+9ogCW/8k50VpTAUMvdlrw3Ya0X9HGmIDhq5OxdwF/FpHZOF00e4B64CfAO6qad6yWq4jcCNwI0KdPHx+V5H+lVbX85t0cXl6xi95donnpunGcMyDV+w+oyIdl7oCvrYRhM5wRNF3PaL+ijTEBx5ug3wP09nid7l7XRFX34rToEZE44FJVLRaRs4BzReQnQBwQISLlqjq32f7zgHkAWVlZeqIH05F8uPEAv1j0NQfLqrj+nH789IKBxER4+Xu13AXLnoCVz7oD/lJ3wA9u36KNMQHJm+RZCQwQkX44AX8ZcIXnBiKSChSqagNwN84IHFT1So9tZgNZzUM+0LjKqrn/rQ28vW4fg7vH89erR5PZO8m7nctdsOyPTsDXVcGw7zsBb/PNGGNOQptBr6p1IjIHeB9neOV8Vd0gIg8A2aq6GDgf+LWIKE7Xzc3tWHOHpKr8Y/UeHvzXRg7V1HPXBQO58bzTiAjzYrbH8oPw+R8he74T8MNnOgGfOqD9CzfGBDxR7Vg9JVlZWZqdne3vMo7L7sJKfv7Gej7bkk9W32QeuXQEp3eNa3vHsgOHA76+GkbMgnPvgtTT279oY0xAEZFVqtriyEa7MvYk1Dcoz32+g99/sJkQgQenD+XKcX3bHg/fUA+f/d551Nc6AX/eXZBy2qkp3BgTVCzoT1DO/lL+5x/rWbu7mEmDu/LQxcPomRTd9o7lLvjn9bB9KQy9BCb9wgLeGNOuLOiPU3VdPU9+spW/LN1GQnQ4f7zsTKZl9vTuwqdvlsFr10JVMXzvCWcmSZtkzBjTzizoj0P2zkL+5x/r2OaqYMbIXtz73SF0ifXiwqeGBmc0zccPQnIGXPW6zUdjjDllLOi9UF5dx2/fy+HFL76hZ2I0z187hvMHdfVu58pCeOMm2PI+DLkYpv0JohLat2BjjPFgQd+G1buKmPPyavaVVnHNWRn897cHHXn3pmPJy4bXZkPZfrjwdzD2BuuqMcacchb0x/De1/u57ZWv6JYQxes3nc3ovsne7ajqzA3/wS8gvgdc9z70Gt2+xRpjTCss6Fvx/Oc7+NW/NpKZnsSz12SREtfCHZ1aUlUCb86BTYth4IVwyVMQ7eUvCGOMaQcW9M00NCi/fncTz3y2gylDuvHEZSOJjgj1bud9a2HhNVC8Cy54CM6aY101xhi/s6D3UFVbz52vreXtdfv44Vl9ue97Qwn15mYgqrDqOXh3LsSkwLXvQJ/x7V+wMcZ4wYLerbiyhhsXrOLLnYXcfeFgbjyvv3dj46vL4V+3w/rX4LTJMGMexB7HVMTGGNPOLOhx5qqZ/dyX7C48xBOXj2RaZk/vdjy4CRb+EAq2wv+7F869E0K8mMTMGGNOoaAP+q/3lHDt8yuprq1nwXVjGd8/xbsd1/wf/OunEBkPVy+C/hPbt1BjjDlBQR30S3IPcvPLq0mOieD/rh/HgG7xbe9UUwnv/jd89RJknAuXPgvx3dq/WGOMOUFBG/R//3IX9y76msHd45k/ewzdEqLa3il/q9NVc3CDM53w+XdDaND+JzTGdBJBl1KqymMfbuZPn2xl4sA0nrxyFHHeXOn69T9g8a0QGgFXvg4DprR/scYY4wNenTkUkakikisiW0XkqFsBikhfEflYRNaJyFIRSXevP1NElovIBvd7s3x9AMejpq6BO19by58+2cqsrN787ZqstkO+vhbevgte/xF0HQI3fWYhb4zpVNpsyopIKPAkMAXIA1aKyGJV3eix2aPAAlV9QUQmAb8GrgYqgR+q6hYR6QmsEpH3VbXY50fShtKqWn7y0mr+szWfO741kFsnn+7d8MmNb8LKZ2D8T2DKAxAa3v7FGmOMD3nTdTMW2Kqq2wFE5BVgOuAZ9EOAn7qXlwCLAFR1c+MGqrpXRA4CacApDfr9JVXMfu5Lth4s53ffH8HMrN7e73zgawgJt5A3xnRa3nTd9AJ2e7zOc6/ztBaY4V6+BIgXkSPGKYrIWCAC2HZipZ6YnP2lXPKXz8krOsT82WOOL+QBXLmQcrqFvDGm0/LV1T13ARNF5CtgIrAHqG98U0R6AC8C16pqQ/OdReRGEckWkWyXy+WjkmDZ1nxmPrWc+gbl1f8az3kD047/Q1w5kDbIZzUZY8yp5k3Q7wE8m8Hp7nVNVHWvqs5Q1ZHAPe51xQAikgC8Ddyjql+09AWqOk9Vs1Q1Ky3tBMK4Bf9cncc1z31Jj6Qo3rh5AkN7Jh7/h9QegqKdkDbYJzUZY4w/eBP0K4EBItJPRCKAy4DFnhuISKqINH7W3cB89/oI4A2cE7Wv+67s1qkqTy7Zyk8XrmV032Reu+lsenlz0+6WFGwFbbAWvTGmU2sz6FW1DpgDvA9sAhaq6gYReUBEprk3Ox/IFZHNQDfgYff6HwDnAbNFZI37caavD6JRXX0DP3/ja373fi7Tz+zJCz8aS2L0SfStu3KdZ2vRG2M6Ma8umFLVd4B3mq37pcfy68BRLXZVfQl46SRr9EpFdR23/P0rPsk5yI/PP43/vmAQId5MMXwsrhyQUEg5zTdFGmOMHwTMlbElh2rJ2VfKgxcP4+rxfX3zoa4c6NIfwry8u5QxxnRAARP0PZOi+ejOicRE+PCQXLnWP2+M6fQCavJ0n4Z8XTUUbLP+eWNMpxdQQe9TBdtA6y3ojTGdngV9a1w5zrN13RhjOjkL+ta4cgGB1AH+rsQYY06KBX1rXDmQnAHhJ3ixlTHGdBAW9K1x5Vr/vDEmIFjQt6S+1pn+wPrnjTEBwIK+JYU7oKHWWvTGmIBgQd8SG3FjjAkgFvQtaZzMLHWgf+swxhgfsKBviSsHEvtAZJy/KzHGmJNmQd8Sm+PGGBNALOiba6iH/M0W9MaYgGFB31zRTqivthE3xpiA4VXQi8hUEckVka0iMreF9/uKyMcisk5ElopIusd714jIFvfjGl8W3y7srlLGmADTZtCLSCjwJHAhMAS4XESGNNvsUZz7wo4AHgB+7d63C3AfMA4YC9wnIsm+K78dNA2ttBE3xpjA4E2LfiywVVW3q2oN8Aowvdk2Q4BP3MtLPN7/NvChqhaqahHwITD15MtuR65ciO8JUYn+rsQYY3zCm6DvBez2eJ3nXudpLTDDvXwJEC8iKV7u27G4cuxErDEmoPjqZOxdwEQR+QqYCOwB6r3dWURuFJFsEcl2uVw+KukENDS4R9xY/7wxJnB4E/R7gN4er9Pd65qo6l5VnaGqI4F73OuKvdnXve08Vc1S1ay0tLTjPAQfKtkNtZXWojfGBBRvgn4lMEBE+olIBHAZsNhzAxFJFZHGz7obmO9efh+4QESS3SdhL3Cv65hsxI0xJgC1GfSqWgfMwQnoTcBCVd0gIg+IyDT3ZucDuSKyGegGPOzetxB4EOeXxUrgAfe6jskmMzPGBKAwbzZS1XeAd5qt+6XH8uvA663sO5/DLfyOzZULsV0hpou/KzHGGJ+xK2M92YgbY0wAsqBvpGq3DzTGBCQL+kale6GmzFr0xpiAY0HfqOlErLXojTGBxYK+kQ2tNMYEKAv6Rq4ciO4Csan+rsQYY3zKgr5R44lYEX9XYowxPmVBD+4RNza00hgTmCzoAcoPQlWx9c8bYwKSBT3Y1AfGmIBmQQ824sYYE9As6MFp0UcmQnx3f1dijDE+Z0EP7hE3g2zEjTEmIFnQg424McYENAv6inyozLf+eWNMwLKgtxOxxpgA51XQi8hUEckVka0iMreF9/uIyBIR+UpE1onIRe714SLygoisF5FNInK3rw/gpDUOrexqQW+MCUxtBr2IhAJPAhcCQ4DLRWRIs83uxbnF4Eice8r+xb1+JhCpqsOB0cB/iUiGb0r3EVcuRMRBQi9/V2KMMe3Cmxb9WGCrqm5X1RrgFWB6s20USHAvJwJ7PdbHikgYEA3UAKUnXbUvNZ6ItRE3xpgA5U3Q9wJ2e7zOc6/zdD9wlYjk4dxb9hb3+teBCmAfsAt4tMPdHNzuKmWMCXC+Ohl7OfC8qqYDFwEvikgIzl8D9UBPoB9wp4j0b76ziNwoItkiku1yuXxUkhcOFUH5fhtaaYwJaN4E/R6gt8frdPc6T9cBCwFUdTkQBaQCVwDvqWqtqh4EPgeymn+Bqs5T1SxVzUpLSzv+ozhRrs3Os7XojTEBzJugXwkMEJF+IhKBc7J1cbNtdgGTAUTkDJygd7nXT3KvjwXGAzm+Kd0HbDIzY0wQaDPoVbUOmAO8D2zCGV2zQUQeEJFp7s3uBG4QkbXA34HZqqo4o3XiRGQDzi+M51R1XXscyAlx5UJYNCT28XclxhjTbsK82UhV38E5yeq57pceyxuBCS3sV44zxLJjcuVA2kAIsevGjDGBK7gTzkbcGGOCQPAGfVUplOZZ/7wxJuAFb9Dnb3GerUVvjAlwwRv0TSNuLOiNMYEtuIM+NBKS+vq7EmOMaVdBHPS5kDoAQr0aeGSMMZ1WEAe93VXKGBMcgjPoayqgeJf1zxtjgkJwBn3+FkCtRW+MCQrBGfR2+0BjTBAJ0qDPgZAw6HLUjMnGGBNwgjTocyHldAgN93clxhjT7oI06G3EjTEmeARf0NdWQdEO6583xgSN4Av6gq2gDdaiN8YEjeALepvjxhgTZLwKehGZKiK5IrJVROa28H4fEVkiIl+JyDoRucjjvREislxENojIehGJ8uUBHDdXLkiIczLWGGOCQJsTvYhIKM4tAacAeRKPp4cAABdmSURBVMBKEVnsvqtUo3txbjH4lIgMwbkbVYaIhAEvAVer6loRSQFqfX4Ux8OV4wyrDIv0axnGGHOqeNOiHwtsVdXtqloDvAJMb7aNAgnu5URgr3v5AmCdqq4FUNUCVa0/+bJPgt1VyhgTZLwJ+l7Abo/Xee51nu4HrhKRPJzW/C3u9QMBFZH3RWS1iPzsJOs9OXU1ULjNTsQaY4KKr07GXg48r6rpwEXAiyISgtM1dA5wpfv5EhGZ3HxnEblRRLJFJNvlcvmopBYUboeGOmvRG2OCijdBvwfo7fE63b3O03XAQgBVXQ5EAak4rf9PVTVfVStxWvujmn+Bqs5T1SxVzUpLSzv+o/BW04gba9EbY4KHN0G/EhggIv1EJAK4DFjcbJtdwGQAETkDJ+hdwPvAcBGJcZ+YnQhsxF9cuYBAygC/lWCMMadam6NuVLVORObghHYoMF9VN4jIA0C2qi4G7gSeEZE7cE7MzlZVBYpE5DGcXxYKvKOqb7fXwbTJlQPJfSEixm8lGGPMqebVffRU9R2cbhfPdb/0WN4ITGhl35dwhlj6n424McYEoeC5Mra+Dgq2WP+8MSboBE/QF+2E+hpr0Rtjgk7wBL2NuDHGBKngC/rUgf6twxhjTrEgCvpcSOwNkfH+rsQYY06pIAp6u6uUMSY4BUfQN9RD/mY7EWuMCUrBEfTFu6Cuylr0xpigFBxBb3eVMsYEseAKehtxY4wJQkES9LkQ3wOik/xdiTHGnHJBEvQ24sYYE7wCP+gbGsBlI26MMcEr8IO+NA9qK6xFb4wJWoEf9K5c59la9MaYIBUEQW9DK40xwc2roBeRqSKSKyJbRWRuC+/3EZElIvKViKwTkYtaeL9cRO7yVeFec+VAbBrEdDnlX22MMR1Bm0EvIqHAk8CFwBDgchEZ0myze4GFqjoS556yf2n2/mPAuydf7gmwu0oZY4KcNy36scBWVd2uqjXAK8D0ZtsokOBeTgT2Nr4hIhcDO4ANJ1/ucVJ1B72diDXGBC9vgr4XsNvjdZ57naf7gatEJA/n3rK3AIhIHPA/wK9OutITUbYPqkutRW+MCWpe3RzcC5cDz6vq70XkLOBFERmG8wvgD6paLiKt7iwiNwI3AvTp08dHJWF3lTLtora2lry8PKqqqvxdiglCUVFRpKenEx4e7vU+3gT9HqC3x+t09zpP1wFTAVR1uYhEAanAOOD7IvJbIAloEJEqVf2z586qOg+YB5CVlaVeV98WG1pp2kFeXh7x8fFkZGRwrAaMMb6mqhQUFJCXl0e/fv283s+brpuVwAAR6SciETgnWxc322YXMBlARM4AogCXqp6rqhmqmgE8Dvxv85BvV64ciE52Rt0Y4yNVVVWkpKRYyJtTTkRISUk57r8m2wx6Va0D5gDvA5twRtdsEJEHRGSae7M7gRtEZC3wd2C2qvquZX6iGkfc2D9I42MW8sZfTuRnz6s+elV9B+ckq+e6X3osbwQmtPEZ9x93dSdDFQ5ugqEXn9KvNcaYjiZwr4ytcEFVsfXPm6AXFxfns89atGgRGzdu9NnnHcvZZ599Qvvdf//9PProoz6upnML3KC3ETfG+NypCPq6ujoAli1b1q7f094aj6Mj8NXwyo7HRtyYU+BXb21g495Sn37mkJ4J3Pe9oa2+P3fuXHr37s3NN98MOC3YsLAwlixZQlFREbW1tTz00ENMn978usaW/eY3v+Gll14iJCSECy+8kEceeYRnnnmGefPmUVNTw+mnn86LL77ImjVrWLx4Mf/+97956KGH+Mc//gHAzTffjMvlIiYmhmeeeYbBgwezbds2rrzySioqKpg+fTqPP/445eXlqCo/+9nPePfddxER7r33XmbNmsXSpUv5xS9+QXJyMjk5OWzevJm4uDjKy8uPq8aYmJg2j7e1/Q4cOMBNN93E9u3bAXjqqac4++yzWbBgAY8++igiwogRI3jxxReZPXs23/3ud/n+978P0FRrS8dx8cUXs3v3bqqqqrjtttu48cYbAXjvvff4+c9/Tn19PampqXz44YcMGjSIZcuWkZaWRkNDAwMHDmT58uWkpZ3cgJIADvociExw7ixlTACZNWsWt99+e1PQL1y4kPfff59bb72VhIQE8vPzGT9+PNOmTWvzxN27777Lm2++yYoVK4iJiaGwsBCAGTNmcMMNNwBw77338uyzz3LLLbcwbdq0IwJu8uTJPP300wwYMIAVK1bwk5/8hE8++YTbbruN2267jcsvv5ynn3666fv++c9/smbNGtauXUt+fj5jxozhvPPOA2D16tV8/fXXRw0bPN4a29LafrfeeisTJ07kjTfeoL6+nvLycjZs2MBDDz3EsmXLSE1NbfruY2l+HPPnz6dLly4cOnSIMWPGcOmll9LQ0MANN9zAp59+Sr9+/SgsLCQkJISrrrqKl19+mdtvv52PPvqIzMzMkw55COigtxE3pv0dq+XdXkaOHMnBgwfZu3cvLpeL5ORkunfvzh133MGnn35KSEgIe/bs4cCBA3Tv3v2Yn/XRRx9x7bXXNrWEu3RxJv/7+uuvuffeeykuLqa8vJxvf/vbR+1bXl7OsmXLmDlzZtO66upqAJYvX86iRYsAuOKKK7jrLmc+w//85z9cfvnlhIaG0q1bNyZOnMjKlStJSEhg7NixLY4NP5kaW9Lafp988gkLFiwAIDQ0lMTERBYsWMDMmTNJTU094ruPpflxPPHEE7zxxhsA7N69my1btuByuTjvvPOatmv83B/96EdMnz6d22+/nfnz53Pttdd6dUxtCeCgz4GBU/1dhTHtYubMmbz++uvs37+fWbNm8fLLL+NyuVi1ahXh4eFkZGSc1JW7s2fPZtGiRWRmZvL888+zdOnSo7ZpaGggKSmJNWvWnMSRHBYbG+vzGn25n6ewsDAaGhoA579DTU1N03uex7F06VI++ugjli9fTkxMDOeff/4x/7/07t2bbt268cknn/Dll1/y8ssvH3dtLQnMk7EVBc6oG+ufNwFq1qxZvPLKK7z++uvMnDmTkpISunbtSnh4OEuWLOGbb77x6nOmTJnCc889R2VlJUBT10RZWRk9evSgtrb2iLCJj4+nrKwMgISEBPr168drr70GOFdtrl27FoDx48c39eG/8sorTfufe+65vPrqq9TX1+Nyufj0008ZO3asT2tsS2v7TZ48maeeegqA+vp6SkpKmDRpEq+99hoFBQVHfHdGRgarVq0CYPHixdTW1rb4XSUlJSQnJxMTE0NOTg5ffPFF03+fTz/9lB07dhzxuQDXX389V111FTNnziQ0NNTr4zqWwAz6fDsRawLb0KFDKSsro1evXvTo0YMrr7yS7Oxshg8fzoIFCxg82Luf/alTpzJt2jSysrI488wzm4YlPvjgg4wbN44JEyYc8VmXXXYZv/vd7xg5ciTbtm3j5Zdf5tlnnyUzM5OhQ4fy5ptvAvD444/z2GOPMWLECLZu3UpiYiIAl1xyCSNGjCAzM5NJkybx29/+ts3upeOtsS2t7ffHP/6RJUuWMHz4cEaPHs3GjRsZOnQo99xzDxMnTiQzM5Of/vSnANxwww38+9//JjMzk+XLl7f618jUqVOpq6vjjDPOYO7cuYwfPx6AtLQ05s2bx4wZM8jMzGTWrFlN+0ybNo3y8nKfddsASEe4gNVTVlaWZmdnn9yHZM+Hf90Bt38NSb3b3t6Y47Bp0ybOOOMMf5fRoVVWVhIdHY2I8Morr/D3v/+96ZeAObbs7GzuuOMOPvvss1a3aelnUERWqWpWS9sHZh+9Kxci4iAx3d+VGBOUVq1axZw5c1BVkpKSmD9/vr9L6hQeeeQRnnrqKZ/1zTcK0KDPgdSBNuLGGLf169dz9dVXH7EuMjKSFStWtMv3nXvuuU399f5y88038/nnnx+x7rbbbvNpl4ivzZ07l7lzj7pb60kL0KDPhf7/z99VGNNhDB8+3GejYzqLJ5980t8ldBiBdzL2ULFzZymb+sAYY4BADPr8zc6zjbgxxhggEIPeJjMzxpgjBGDQ50JYNCT58N6zxhjTiXkV9CIyVURyRWSriBx1SlhE+ojIEhH5SkTWichF7vVTRGSViKx3P0/y9QEcxZUDqQMgxDdXlBnTmS1duvSUTfd70UUXUVxcfNz7Pf/888yZM6cdKjKN2hx1IyKhwJPAFCAPWCkii913lWp0L84tBp8SkSE4d6PKAPKB76nqXhEZhnM7wl4+PoYjuXKhz1nt+hXGdBZLly4lLi7uhG/i4Q1VRVV555132t64A2s8jpCQwOvo8OaIxgJbVXW7qtYArwDNJ7pWIMG9nAjsBVDVr1R1r3v9BiBaRCJPvuxWVJdByW7rnzenzrtz4bnv+PbxbtvjqBcsWNA0lcDVV1/NW2+9xbhx4xg5ciTf+ta3OHDgADt37uTpp5/mD3/4A2eeeSafffYZLpeLSy+9lDFjxjBmzJimceYul4spU6YwdOhQrr/+evr27Ut+fj4Ajz32GMOGDWPYsGE8/vjjAOzcuZNBgwbxwx/+kGHDhrF7924yMjKa9mleH9Bijd5obb/GaQKGDx/OiBEjmubWee+99xg1ahSZmZlMnjwZOPquU8OGDWPnzp0tHsePf/xjsrKyGDp0KPfdd1/TPitXruTss88mMzOTsWPHUlZWxnnnnXfEsNVzzjnH79cPtMSbcfS9gN0er/OAcc22uR/4QERuAWKBb7XwOZcCq1W1+gTq9I6NuDFBoKU50kWEL774AhHhb3/7G7/97W/5/e9/z0033URcXFzTNMFXXHEFd9xxB+eccw67du3i29/+Nps2beJXv/oVkyZN4u677+a9997j2WefBZwrXJ977jlWrFiBqjJu3DgmTpxIcnIyW7Zs4YUXXmiav+VY9YETgi3V2JbW9nvwwQdJTExk/fr1ABQVFeFyuY6a570tzY/j4YcfpkuXLtTX1zN58mTWrVvH4MGDmTVrFq+++ipjxoyhtLSU6OhorrvuOp5//nkef/xxNm/eTFVVFZmZmd7/zzxFfHXB1OXA86r6exE5C3hRRIapagOAiAwFfgNc0NLOInIjcCNAnz4ncRLV7iplTrULHznlX/nJJ58cNUf6+vXrmTVrFvv27aOmpqbFed3Bmdvd81aApaWllJeX85///KdpzvSpU6eSnJwMOPPHX3LJJU2Tds2YMYPPPvuMadOm0bdv36NCvrX6APLy8ryqsbnW9vvoo4+OmBkzOTmZt956q8V53o+l+XEsXLiQefPmUVdXx759+9i4cSMiQo8ePRgzZgzgzNwJznTRDz74IL/73e+YP38+s2fP9uqYTjVvum72AJ4zg6W713m6DlgIoKrLgSggFUBE0oE3gB+q6raWvkBV56lqlqpmndTdVFw5EBoByRkn/hnGdEK33HILc+bMYf369fz1r39tdc7zhoYGvvjiC9asWcOaNWvYs2fPCd88/Hjnj/e2Rl/t58lz/njgiM/wPI4dO3bw6KOP8vHHH7Nu3Tq+853vHPP7YmJimDJlCm+++SYLFy7kyiuvPO7aTgVvgn4lMEBE+olIBHAZsLjZNruAyQAicgZO0LtEJAl4G5irqp/T3ly5kDIAQgNzZgdjgBbnSC8pKaFXL2ecwwsvvNC0ref88QAXXHABf/rTn5peN/YvT5gwgYULFwLwwQcfUFRUBDhz1ixatIjKykoqKip44403OPfcc4+7PqDVGtvS2n5Tpkw5YpqDoqKiVud5z8jIYPXq1YBzq7/G95srLS0lNjaWxMREDhw4wLvvvgvAoEGD2LdvHytXrgScOe0bb/59/fXXc+uttzJmzJimv4Q6mjaDXlXrgDk4I2Y24Yyu2SAiD4jINPdmdwI3iMha4O/AbHXmP54DnA78UkTWuB9d2+VIwGnR24lYE+BamiP9/vvvZ+bMmYwePbqpywTge9/7Hm+88UbTydgnnniC7OxsRowYwZAhQ5ru53rffffxwQcfMGzYMF577TW6d+9OfHw8o0aNYvbs2YwdO5Zx48Zx/fXXM3LkyOOuD2i1xra0tt+9995LUVERw4YNIzMzkyVLlrQ6z/ull15KYWEhQ4cO5c9//jMDBw5s8bsyMzMZOXIkgwcP5oorrmDChAkARERE8Oqrr3LLLbeQmZnJlClTmlr6o0ePJiEhoUNPlhY489HXVML/9oTz74bz/8f3hRnjFojz0VdXVxMaGkpYWBjLly/nxz/+cdBNgnai9u7dy/nnn09OTs4pG5oZvPPR11TAsEuh97FvS2aMOdquXbv4wQ9+QENDAxERETzzzDP+LqlTWLBgAffccw+PPfZYhx5/HzgtemNOkUBs0XcEDz/8cNP9ZxvNnDmTe+65x08VdVzB26I3xnRq99xzj4V6O+m4f2sY04F1tL+ETfA4kZ89C3pjjlNUVBQFBQUW9uaUU1UKCgqIioo6rv2s68aY45Senk5eXh4ul8vfpZggFBUVRXp6+nHtY0FvzHEKDw/3+vJ9YzoC67oxxpgAZ0FvjDEBzoLeGGMCXIe7YEpEXMA3J/ERqTh3tuoMOlOt0Lnq7Uy1QueqtzPVCp2r3pOpta+qtjj9b4cL+pMlItmtXR3W0XSmWqFz1duZaoXOVW9nqhU6V73tVat13RhjTICzoDfGmAAXiEE/z98FHIfOVCt0rno7U63QuertTLVC56q3XWoNuD56Y4wxRwrEFr0xxhgPARP0IjJVRHJFZKuIzPV3PcciIr1FZImIbBSRDSJym79raouIhIrIVyLyL3/X0hYRSRKR10UkR0Q2ichZ/q6pNSJyh/tn4GsR+buIHN9sVe1MROaLyEER+dpjXRcR+VBEtrifO8SNUlup9Xfun4N1IvKG+z7WHUJL9Xq8d6eIqIh4f8/FYwiIoBeRUOBJ4EJgCHC5iAzxb1XHVAfcqapDgPHAzR28XoDbcO4Z3Bn8EXhPVQcDmXTQukWkF3ArkKWqw4BQ4DL/VnWU54GpzdbNBT5W1QHAx+7XHcHzHF3rh8AwVR0BbAbuPtVFHcPzHF0vItIbuADY5asvCoigB8YCW1V1u6rWAK8A0/1cU6tUdZ+qrnYvl+EEUS//VtU6EUkHvgP8zd+1tEVEEoHzgGcBVLVGVYv9W9UxhQHRIhIGxAB7/VzPEVT1U6Cw2erpwAvu5ReAi09pUa1oqVZV/UBV69wvvwCOb9rHdtTKf1uAPwA/A3x2AjVQgr4XsNvjdR4dODg9iUgGMBJY4d9KjulxnB+8Bn8X4oV+gAt4zt3V9DcRifV3US1R1T3Aozgtt31Aiap+4N+qvNJNVfe5l/cD3fxZzHH4EfCuv4s4FhGZDuxR1bW+/NxACfpOSUTigH8At6tqqb/raYmIfBc4qKqr/F2Ll8KAUcBTqjoSqKDjdC0cwd23PR3nl1NPIFZErvJvVcdHnWF7HX7onojcg9Nl+rK/a2mNiMQAPwd+6evPDpSg3wP09nid7l7XYYlIOE7Iv6yq//R3PccwAZgmIjtxusQmichL/i3pmPKAPFVt/AvpdZzg74i+BexQVZeq1gL/BM72c03eOCAiPQDczwf9XM8xichs4LvAldqxx5OfhvNLf63731s6sFpEup/sBwdK0K8EBohIPxGJwDmhtdjPNbVKRASnD3mTqj7m73qORVXvVtV0Vc3A+e/6iap22Fanqu4HdovIIPeqycBGP5Z0LLuA8SIS4/6ZmEwHPXHczGLgGvfyNcCbfqzlmERkKk634zRVrfR3PceiqutVtauqZrj/veUBo9w/0yclIILefbJlDvA+zj+Uhaq6wb9VHdME4Gqc1vEa9+MifxcVQG4BXhaRdcCZwP/6uZ4Wuf/qeB1YDazH+ffYoa7iFJG/A8uBQSKSJyLXAY8AU0RkC85fJY/4s8ZGrdT6ZyAe+ND97+xpvxbpoZV62+e7OvZfMsYYY05WQLTojTHGtM6C3hhjApwFvTHGBDgLemOMCXAW9MYYE+As6I0xJsBZ0BtjTICzoDfGmAD3/wG3QFFeIyJPNwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "exp = get_H()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def dz_dw_flatten(z, params):\n",
    "    \"\"\" Calculate dz/dparams and flatten the result \"\"\"\n",
    "    return tf.concat([tf.reshape(x, shape = (-1,)) for x in tf.gradients(z, params)], axis = 0)\n",
    "\n",
    "def iterate_flatten(tensor):\n",
    "    \"\"\" Iterate over flattened items of a tensor \"\"\"\n",
    "    if type(tensor) == list:\n",
    "        for t in tensor:\n",
    "            for v in iterate_flatten(t):\n",
    "                yield v\n",
    "    elif len(tensor.shape) == 0:\n",
    "        yield tensor\n",
    "    else:\n",
    "        for idx in range(tensor.shape[0]):\n",
    "            for v in iterate_flatten(tensor[idx]):\n",
    "                yield v\n",
    "\n",
    "def hessian(loss, params):\n",
    "    \"\"\" Compute a hessian of the loss w.r.t. params \"\"\"\n",
    "    with tf.name_scope('hessian'):\n",
    "        grads = tf.gradients(loss, params)\n",
    "        grad_components = list(iterate_flatten(grads))\n",
    "        hessian = [dz_dw_flatten(t, params) for t in tqdm(grad_components)]\n",
    "        return hessian\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 1574/84060 [36:18<46:39:53,  2.04s/it]"
     ]
    }
   ],
   "source": [
    "hessian(exp.loss, exp.model_correct.weights)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tf.Variable 'dense_11/kernel:0' shape=(784, 200) dtype=float32_ref>,\n",
       " <tf.Variable 'dense_11/bias:0' shape=(200,) dtype=float32_ref>,\n",
       " <tf.Variable 'dense_12/kernel:0' shape=(200, 100) dtype=float32_ref>,\n",
       " <tf.Variable 'dense_12/bias:0' shape=(100,) dtype=float32_ref>,\n",
       " <tf.Variable 'dense_13/kernel:0' shape=(100, 100) dtype=float32_ref>,\n",
       " <tf.Variable 'dense_13/bias:0' shape=(100,) dtype=float32_ref>,\n",
       " <tf.Variable 'dense_14/kernel:0' shape=(100, 100) dtype=float32_ref>,\n",
       " <tf.Variable 'dense_14/bias:0' shape=(100,) dtype=float32_ref>,\n",
       " <tf.Variable 'dense_15/kernel:0' shape=(100, 10) dtype=float32_ref>,\n",
       " <tf.Variable 'dense_15/bias:0' shape=(10,) dtype=float32_ref>]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "exp.model_correct.weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total runs: 15\n"
     ]
    }
   ],
   "source": [
    "# all scales of the network (up and down)\n",
    "scales = np.linspace(0.5, 5, 5)\n",
    "\n",
    "# repetitions for each scale\n",
    "repetitions = 3\n",
    "\n",
    "print('Total runs: %d' % (len(scales) * repetitions))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "def process_scale(s):\n",
    "    \"\"\" Run many repetitions for one scale \"\"\"\n",
    "    return [get_W(scale = s, epochs = 0) for _ in range(repetitions)]\n",
    "\n",
    "def process_all():\n",
    "    \"\"\" Random variable with all results \"\"\"\n",
    "    return {scale: process_scale(scale) for scale in tqdm(scales)}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "  0%|          | 0/5 [00:00<?, ?it/s]\u001b[A\n",
      " 20%|██        | 1/5 [00:15<01:01, 15.42s/it]\u001b[A\n",
      " 40%|████      | 2/5 [00:32<00:47, 15.84s/it]\u001b[A\n",
      " 60%|██████    | 3/5 [00:51<00:33, 16.83s/it]\u001b[A\n",
      " 80%|████████  | 4/5 [01:12<00:18, 18.15s/it]\u001b[A\n",
      "100%|██████████| 5/5 [01:36<00:00, 19.83s/it]\u001b[A"
     ]
    }
   ],
   "source": [
    "# obtaining the results\n",
    "results = process_all()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYMAAAD4CAYAAAAO9oqkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAS2klEQVR4nO3dcayd9X3f8fenpjCWxLOJPZfYXk1bL5KDOie5JURtV1oY2EyLqYYysrW4EarbAVqrdVJpM4kqaaQwKamCStncxYqZ2hKaNsVKTahxQFmkmnBDPYOdETsumW0cc2NjnAaWzOl3f5yfycG+1/fce88999h+v6Sj+zzf5/c8z/dcX9/PfZ7znOekqpAkXdh+YK4bkCTNPcNAkmQYSJIMA0kShoEkCbhorhuYrkWLFtWKFSvmug1JOqd8+ctf/mZVLT69fs6GwYoVKxgdHZ3rNiTpnJLk6+PVPU0kSTIMJEmGgSQJw0CShGEgScIwkCRhGEiSMAwkSRgGkiTO4XcgS5N67pG52e9b187NfqUZ8MhAkmQYSJIMA0kShoEkiR7CIMnyJI8n2ZNkd5Jfa/XfSXIoyc72uLFrnd9Ksi/Jc0lu6KqvabV9Se7qql+R5MlW/1SSi/v9RCVJE+vlyOAk8BtVtQq4Grgjyaq27PeqanV7bAVoy24B3gasAf4gybwk84D7gLXAKuB9Xdu5p23rx4CXgNv69PwkST2YNAyq6nBVPd2mvwV8BVh6llXWAQ9W1Xeq6m+BfcBV7bGvqvZX1XeBB4F1SQL8HPDptv5m4KbpPiFJ0tRN6TWDJCuAtwNPttKdSXYl2ZRkYastBQ50rXaw1Saqvxk4XlUnT6tLkgak5zBI8kbgz4Bfr6oTwP3AjwKrgcPAR2elw9f3sCHJaJLRsbGx2d6dJF0wenoHcpIfpBMEf1RVfw5QVUe6lv8h8Nk2ewhY3rX6slZjgvpRYEGSi9rRQff416mqjcBGgJGRkeqld6lfdh443tO4b37vyOSDenTdqiV925Z0Nr1cTRTgE8BXqupjXfXLu4b9PPBsm94C3JLkkiRXACuBLwFPASvblUMX03mReUtVFfA4cHNbfz3w8MyeliRpKno5MvhJ4BeBZ5LsbLXfpnM10GqggOeBXwGoqt1JHgL20LkS6Y6q+h5AkjuBR4F5wKaq2t2295vAg0l+F/gbOuEjSRqQScOgqr4IZJxFW8+yzoeBD49T3zreelW1n87VRpKkOeA7kCVJhoEkyTCQJGEYSJIwDCRJGAaSJAwDSRKGgSQJw0CShGEgSaLHu5ZK6t2iFz7fv43NWzC18W9d279964LikYEkySMDaTqefnXvQPazmp8YyH4kjwwkSYaBJMkwkCRhGEiSMAwkSRgGkiQMA0kShoEkCcNAkoRhIEnC21HoHPbYniOvm3/2+F+/bn7+0V2DbEc6p3lkIEkyDCRJhoEkCcNAkoRhIEnCMJAkYRhIkughDJIsT/J4kj1Jdif5tVa/LMm2JHvb14WtniT3JtmXZFeSd3Rta30bvzfJ+q76O5M809a5N0lm48lKksbXy5HBSeA3qmoVcDVwR5JVwF3A9qpaCWxv8wBrgZXtsQG4HzrhAdwNvAu4Crj7VIC0Mb/ctd6amT81SVKvJg2DqjpcVU+36W8BXwGWAuuAzW3YZuCmNr0OeKA6dgALklwO3ABsq6pjVfUSsA1Y05bNr6odVVXAA13bkiQNwJReM0iyAng78CSwpKoOt0XfAJa06aXAga7VDrba2eoHx6mPt/8NSUaTjI6NjU2ldUnSWfQcBkneCPwZ8OtVdaJ7WfuLvvrc2xmqamNVjVTVyOLFi2d7d5J0wegpDJL8IJ0g+KOq+vNWPtJO8dC+vtjqh4DlXasva7Wz1ZeNU5ckDUgvVxMF+ATwlar6WNeiLcCpK4LWAw931W9tVxVdDbzcTic9ClyfZGF74fh64NG27ESSq9u+bu3aliRpAHq5hfVPAr8IPJNkZ6v9NvAR4KEktwFfB97blm0FbgT2Aa8A7weoqmNJPgQ81cZ9sKqOtenbgU8ClwKPtIckaUAmDYOq+iIw0XX/144zvoA7JtjWJmDTOPVR4MrJepEkzQ7fgSxJMgwkSYaBJAnDQJKEYSBJwjCQJGEYSJIwDCRJGAaSJAwDSRKGgSQJw0CShGEgScIwkCRhGEiSMAwkSRgGkiQMA0kShoEkCcNAkoRhIEnCMJAkYRhIkjAMJEkYBpIkDANJEoaBJAnDQJKEYSBJwjCQJGEYSJLoIQySbEryYpJnu2q/k+RQkp3tcWPXst9Ksi/Jc0lu6KqvabV9Se7qql+R5MlW/1SSi/v5BCVJk+vlyOCTwJpx6r9XVavbYytAklXALcDb2jp/kGReknnAfcBaYBXwvjYW4J62rR8DXgJum8kTkiRN3aRhUFVfAI71uL11wINV9Z2q+ltgH3BVe+yrqv1V9V3gQWBdkgA/B3y6rb8ZuGmKz0GSNEMzec3gziS72mmkha22FDjQNeZgq01UfzNwvKpOnlYfV5INSUaTjI6Njc2gdUlSt+mGwf3AjwKrgcPAR/vW0VlU1caqGqmqkcWLFw9il5J0QbhoOitV1ZFT00n+EPhsmz0ELO8auqzVmKB+FFiQ5KJ2dNA9XpI0INM6MkhyedfszwOnrjTaAtyS5JIkVwArgS8BTwEr25VDF9N5kXlLVRXwOHBzW3898PB0epIkTd+kRwZJ/gS4BliU5CBwN3BNktVAAc8DvwJQVbuTPATsAU4Cd1TV99p27gQeBeYBm6pqd9vFbwIPJvld4G+AT/Tt2UmSejJpGFTV+8YpT/gLu6o+DHx4nPpWYOs49f10rjaSJM0R34EsSTIMJEmGgSSJaV5aKmkwdh44PqXx3/zekckHncV1q5bMaH2duzwykCQZBpIkw0CShGEgScIwkCTh1UTSUHv61b1TGn/i+KXT2s+VC949rfV0/vDIQJJkGEiSPE2k2fbcI7O26UUvvP4NWfOneEpF0vd5ZCBJMgwkSYaBJAnDQJKEYSBJwjCQJGEYSJIwDCRJGAaSJAwDSRKGgSQJw0CShGEgScIwkCRhGEiSMAwkSfjhNhqAJ17aMyvb3f/qt2dlu9KFaNIjgySbkryY5Nmu2mVJtiXZ274ubPUkuTfJviS7kryja531bfzeJOu76u9M8kxb594k6feTlCSdXS+niT4JrDmtdhewvapWAtvbPMBaYGV7bADuh054AHcD7wKuAu4+FSBtzC93rXf6viRJs2zSMKiqLwDHTiuvAza36c3ATV31B6pjB7AgyeXADcC2qjpWVS8B24A1bdn8qtpRVQU80LUtSdKATPcF5CVVdbhNfwNY0qaXAge6xh1stbPVD45TlyQN0IyvJmp/0VcfeplUkg1JRpOMjo2NDWKXknRBmG4YHGmneGhfX2z1Q8DyrnHLWu1s9WXj1MdVVRuraqSqRhYvXjzN1iVJp5tuGGwBTl0RtB54uKt+a7uq6Grg5XY66VHg+iQL2wvH1wOPtmUnklzdriK6tWtbkqQBmfR9Bkn+BLgGWJTkIJ2rgj4CPJTkNuDrwHvb8K3AjcA+4BXg/QBVdSzJh4Cn2rgPVtWpF6Vvp3PF0qXAI+0hSRqgScOgqt43waJrxxlbwB0TbGcTsGmc+ihw5WR9SJJmj7ejkCQZBpIk702kPnpsz5EzaoteOO49hAZo/tFd01pv0SuvdibmLZjejt+6dnrraWh4ZCBJMgwkSYaBJAnDQJKEYSBJwjCQJGEYSJIwDCRJGAaSJAwDSRKGgSQJw0CShGEgScIwkCRhGEiSMAwkSRgGkiQMA0kShoEkCcNAkoRhIEnCMJAkYRhIkjAMJEkYBpIkDANJEoaBJAnDQJLEDMMgyfNJnkmyM8loq12WZFuSve3rwlZPknuT7EuyK8k7urazvo3fm2T9zJ6SJGmq+nFk8LNVtbqqRtr8XcD2qloJbG/zAGuBle2xAbgfOuEB3A28C7gKuPtUgEiSBmM2ThOtAza36c3ATV31B6pjB7AgyeXADcC2qjpWVS8B24A1s9CXJGkCMw2DAv4qyZeTbGi1JVV1uE1/A1jSppcCB7rWPdhqE9XPkGRDktEko2NjYzNsXZJ0ykUzXP+nqupQkn8MbEvyv7sXVlUlqRnuo3t7G4GNACMjI33briRd6GZ0ZFBVh9rXF4HP0Dnnf6Sd/qF9fbENPwQs71p9WatNVJckDci0wyDJG5K86dQ0cD3wLLAFOHVF0Hrg4Ta9Bbi1XVV0NfByO530KHB9koXthePrW02SNCAzOU20BPhMklPb+eOq+lySp4CHktwGfB14bxu/FbgR2Ae8ArwfoKqOJfkQ8FQb98GqOjaDviRJUzTtMKiq/cA/G6d+FLh2nHoBd0ywrU3Apun2IkmaGd+BLEkyDCRJhoEkiZm/z0DSeeDpV/d2vn51euufOPqdnsdeueDdr01ft2rJWUZqkDwykCQZBpIkw0CShGEgScIwkCRhGEiSMAwkSfg+gwvaEwee6Ov2nj1+/Iza/Hb9uqThZhhcKJ575MzaS3v6uov5R7/d1+1JGhxPE0mSDANJkmEgScIwkCRhGEiSMAwkSXhp6XntsT1HXpte9MKZ7wHY/6qXgkrq8MhAkmQYSJI8TSSpD+Yf3dXz2EWvvPr9mXkLZrbjt66d2fp6jUcGkiTDQJJkGEiSMAwkSRgGkiQMA0kSXloqacCe7vr0u+MvvWFmGztw6YSLrll+zcy2fYExDCTNmf1jM7slyom/P/M2K6ec/NaRM2rXrVoyo/2dz4YmDJKsAT4OzAP+e1V9ZI5bmhP9/Fzi7s8k9rOIJZ3NUIRBknnAfcC/AA4CTyXZUlX9/ZDeOfTYniMseuHzk47b38df2vP7tiVJ57uhCAPgKmBfVe0HSPIgsA44b8JAUv+d7TYY/2ecZU+cmOFrFN2WvvO1yfPh9YlhCYOlwIGu+YPAu04flGQDsKHN/l2S53rc/iLgmzPqcHbY19TY19QMY1/D2BNcWH398HjFYQmDnlTVRmDjVNdLMlpVI7PQ0ozY19TY19QMY1/D2BPYFwzP+wwOAcu75pe1miRpAIYlDJ4CVia5IsnFwC3AljnuSZIuGENxmqiqTia5E3iUzqWlm6pqdx93MeVTSwNiX1NjX1MzjH0NY09gX6SqBrUvSdKQGpbTRJKkOWQYSJLOzzBIclmSbUn2tq8LzzJ2fpKDSX5/GPpK8sNJnk6yM8nuJL86JH2tTvLXraddSf7NMPTVxn0uyfEkn53FXtYkeS7JviR3jbP8kiSfasufTLJitnqZYl//vP08nUxy8yB66rGv/5hkT/tZ2p5k3Gvf56CvX03yTPv/98Ukq4ahr65x/zpJJen/5aZVdd49gP8C3NWm7wLuOcvYjwN/DPz+MPQFXAxc0qbfCDwPvGUI+vqnwMo2/RbgMLBgrvtqy64F/hXw2VnqYx7wNeBH2r/P/wJWnTbmduC/tulbgE8N4Oepl75WAD8OPADcPNs9TaGvnwX+YZv+90P0/ZrfNf0e4HPD0Fcb9ybgC8AOYKTffZyXRwZ0bmWxuU1vBm4ab1CSdwJLgL8alr6q6rtV9Z02ewmDOXrrpa+vVtXeNv0C8CKweK77av1sB741i328druUqvoucOp2Kd26e/00cG2SzGJPPfVVVc9X1S7g72e5l6n29XhVvdJmd9B5b9Ew9HWia/YNwCCusOnl5wvgQ8A9wP+djSbO1zBYUlWH2/Q36PzCf50kPwB8FPhPw9QXQJLlSXbRuUXHPe2X75z31dXfVXT+gvnaMPU1i8a7XcrSicZU1UngZeDNQ9DXXJhqX7cBj8xqRx099ZXkjiRfo3Nk+h+Goa8k7wCWV9VfzlYTQ/E+g+lI8hjwQ+Ms+kD3TFVVkvHS/XZga1Ud7OcfcH3oi6o6APx4krcAf5Hk01V15s3ZB9xX287lwP8A1lfVjP/a7FdfOjcl+QVgBPiZue7llKq6D7gvyb8F/jOwfi77aX+4fgz4pdnczzkbBlV13UTLkhxJcnlVHW6/vF4cZ9i7gZ9Ocjudc/MXJ/m7qprwxZsB9dW9rReSPAv8NJ1TD3PaV5L5wF8CH6iqHTPpp599DUAvt0s5NeZgkouAfwQcHYK+5kJPfSW5jk7o/0zXqdE576vLg8D9s9pRx2R9vQm4Enii/eH6Q8CWJO+pqtF+NXG+nibawvfTfD3w8OkDqurfVdU/qaoVdE4VPTDTIOhHX0mWJbm0TS8Efgro9e6ss9nXxcBn6HyfZhRM/exrQHq5XUp3rzcDn6/2qt8c9zUXJu0ryduB/wa8p6oGFfK99LWya/ZfAoP4VKiz9lVVL1fVoqpa0X5f7aDzfetbEJza0Xn3oHOudjudf8jHgMtafYTOp6idPv6XGMzVRJP2RecDfnbRuaJgF7BhSPr6BeD/ATu7Hqvnuq82/z+BMeBVOudbb5iFXm4EvkrndZIPtNoH6fynBPgHwJ8C+4AvAT8y2/9uPfb1E+178m06Ryq7h6Svx4AjXT9LW4akr48Du1tPjwNvG4a+Thv7BLNwNZG3o5AknbeniSRJU2AYSJIMA0mSYSBJwjCQJGEYSJIwDCRJwP8HmWD32kgyH+sAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# untrained\n",
    "for scale in scales:\n",
    "    plt.figure()\n",
    "    for i in range(repetitions):\n",
    "        plt.hist(results[scale][i].flatten(), alpha = 0.3)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD4CAYAAAAXUaZHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAV6klEQVR4nO3dcZBd5X3e8e8TETAmxRJoq2JJWHis4jLuGNMdgutOQwA7QDsWnWIqt6kVqo4yKUmTup0a13847TRTu9MptScdUo1JLNLUmJB4UFMcBwRMpzOGWGAsgwjWWg6RBEhrWQLbYGLsX/+479oXsbv3rvbelTh8PzN37jnv+55zfnv26tmjc889N1WFJKlbfuJEFyBJGj3DXZI6yHCXpA4y3CWpgwx3SeqgU050AQArV66sdevWnegyJOlV5aGHHvpmVU3M1ndShPu6devYuXPniS5Dkl5Vkjw5V5+nZSSpgwx3Seogw12SOshwl6QOMtwlqYMMd0nqoKHCPcm/SvJYkkeTfCbJ65Kcl+TBJFNJPpvk1Db2tDY/1frXjfMHkCS90sBwT7Ia+JfAZFW9DVgGbAQ+DtxUVW8BjgCb2yKbgSOt/aY2TpK0hIY9LXMKcHqSU4DXA08DlwF3tP5twDVtekObp/VfniSjKVeSNIyBn1CtqgNJ/gvwF8ALwJ8ADwFHq+qlNmw/sLpNrwb2tWVfSvIscDbwzf71JtkCbAE499xzF/+TSItwz+6DS77NKy5YteTb1GvHMKdlVtA7Gj8PeCNwBnDlYjdcVVurarKqJicmZr01giTpOA1zWuYK4BtVNV1V3wf+EHgXsLydpgFYAxxo0weAtQCt/w3A4ZFWLUma1zDh/hfAJUle386dXw7sBu4Drm1jNgF3tuntbZ7Wf2/5Ra2StKQGhntVPUjvjdGHga+2ZbYCHwI+mGSK3jn1W9oitwBnt/YPAjeOoW5J0jyGuuVvVX0U+OgxzXuBi2cZ+z3gfYsvTZJ0vPyEqiR1kOEuSR1kuEtSBxnuktRBhrskdZDhLkkdZLhLUgcZ7pLUQUN9iEnS6HknSo2TR+6S1EEeuUvNo0e/uKjlzzy8a6hxF52+flHbAfjmGy9b9DrUbR65S1IHGe6S1EGGuyR1kOEuSR1kuEtSBxnuktRBA8M9yflJHul7PJfk15KcleTuJHva84o2Pkk+mWQqya4kF43/x5Ak9RvmO1SfqKoLq+pC4G8BzwOfo/fdqDuqaj2wgx9/V+pVwPr22ALcPI7CJUlzW+hpmcuBr1fVk8AGYFtr3wZc06Y3ALdWzwPA8iTnjKRaSdJQFhruG4HPtOlVVfV0m34GmLlpxWpgX98y+1vbyyTZkmRnkp3T09MLLEOSNJ+hwz3JqcB7gd8/tq+qCqiFbLiqtlbVZFVNTkxMLGRRSdIACzlyvwp4uKpmbmV3cOZ0S3s+1NoPAGv7llvT2iRJS2Qh4f5+fnxKBmA7sKlNbwLu7Gv/QLtq5hLg2b7TN5KkJTDUXSGTnAG8G/jFvuaPAbcn2Qw8CVzX2u8Crgam6F1Zc/3IqpU64OEX9ix6Hc8dPX3gmLctf+eit6NXr6HCvaq+C5x9TNthelfPHDu2gBtGUp0k6bj4CVVJ6iDDXZI6yHCXpA4y3CWpgwx3Seogw12SOshwl6QOMtwlqYMMd0nqIMNdkjrIcJekDjLcJamDhrpxmPSq9sTnBw5Z+dRRzhzB3Rqlk4VH7pLUQYa7JHWQ4S5JHWS4S1IHDRXuSZYnuSPJnyV5PMk7k5yV5O4ke9rzijY2ST6ZZCrJriQXjfdHkCQda9gj908Af1xVbwXeDjwO3AjsqKr1wI42D3AVsL49tgA3j7RiSdJAA8M9yRuAvwvcAlBVf1lVR4ENwLY2bBtwTZveANxaPQ8Ay5OcM/LKJUlzGubI/TxgGvidJF9O8qkkZwCrqurpNuYZYFWbXg3s61t+f2t7mSRbkuxMsnN6evr4fwJJ0isME+6nABcBN1fVO4Dv8uNTMABUVQG1kA1X1daqmqyqyYmJiYUsKkkaYJhw3w/sr6oH2/wd9ML+4MzplvZ8qPUfANb2Lb+mtUmSlsjAcK+qZ4B9Sc5vTZcDu4HtwKbWtgm4s01vBz7Qrpq5BHi27/SNJGkJDHtvmV8Bfi/JqcBe4Hp6fxhuT7IZeBK4ro29C7gamAKeb2MlSUtoqHCvqkeAyVm6Lp9lbAE3LLIuSdIi+AlVSeogw12SOshwl6QOMtwlqYMMd0nqIMNdkjrIcJekDjLcJamDDHdJ6iDDXZI6yHCXpA4a9sZhkk4iZx7eNXDMyudfeGXjsuUL39j5Vy18GZ1wHrlLUgcZ7pLUQYa7JHWQ4S5JHWS4S1IHDRXuSf48yVeTPJJkZ2s7K8ndSfa05xWtPUk+mWQqya4kF43zB5AkvdJCjtx/tqourKqZr9u7EdhRVeuBHW0e4CpgfXtsAW4eVbGSpOEs5rTMBmBbm94GXNPXfmv1PAAsT3LOIrYjSVqgYcO9gD9J8lCSLa1tVVU93aafAVa16dXAvr5l97e2l0myJcnOJDunp6ePo3RJ0lyG/YTq36mqA0n+KnB3kj/r76yqSlIL2XBVbQW2AkxOTi5oWUnS/IY6cq+qA+35EPA54GLg4MzplvZ8qA0/AKztW3xNa5MkLZGB4Z7kjCR/ZWYaeA/wKLAd2NSGbQLubNPbgQ+0q2YuAZ7tO30jSVoCw5yWWQV8LsnM+P9VVX+c5EvA7Uk2A08C17XxdwFXA1PA88D1I69akjSvgeFeVXuBt8/Sfhi4fJb2Am4YSXWSpOPiJ1QlqYMMd0nqIMNdkjrIcJekDjLcJamDDHdJ6iDDXZI6yHCXpA4y3CWpgwx3Seogw12SOshwl6QOMtwlqYMMd0nqIMNdkjrIcJekDho63JMsS/LlJH/U5s9L8mCSqSSfTXJqaz+tzU+1/nXjKV2SNJeFHLn/KvB43/zHgZuq6i3AEWBza98MHGntN7VxkqQlNMx3qJJkDfD3gN8APpjeF6peBvzjNmQb8OvAzcCGNg1wB/CbSdK+fk8a6J7dB0e6vpVPHR3p+qRXg2GP3P8b8G+BH7b5s4GjVfVSm98PrG7Tq4F9AK3/2TZekrREBh65J/n7wKGqeijJpaPacJItwBaAc889d1SrVQc9evSLi1r+zBf2jKgS6dVjmCP3dwHvTfLnwG30Tsd8AlieZOaPwxrgQJs+AKwFaP1vAA4fu9Kq2lpVk1U1OTExsagfQpL0cgPDvao+XFVrqmodsBG4t6r+CXAfcG0btgm4s01vb/O0/ns93y5JS2sx17l/iN6bq1P0zqnf0tpvAc5u7R8EblxciZKkhRrqapkZVXU/cH+b3gtcPMuY7wHvG0FtkqTj5CdUJamDDHdJ6qAFnZaR9Orx8CyXgB49csbCV7Tv9IFDLl176cLXq7HyyF2SOshwl6QOMtwlqYMMd0nqIMNdkjrIcJekDjLcJamDDHdJ6iDDXZI6yHCXpA4y3CWpgwx3SeogbxwmvYbsnf7ugpd57odHB4556dsH5+y74oJVC96mFs8jd0nqoIHhnuR1Sf40yVeSPJbk37f285I8mGQqyWeTnNraT2vzU61/3Xh/BEnSsYY5cn8RuKyq3g5cCFyZ5BLg48BNVfUW4AiwuY3fDBxp7Te1cZKkJTQw3KvnO232J9ujgMuAO1r7NuCaNr2hzdP6L0+SkVUsSRpoqHPuSZYleQQ4BNwNfB04WlUvtSH7gdVtejWwD6D1PwucPcs6tyTZmWTn9PT04n4KSdLLDBXuVfWDqroQWANcDLx1sRuuqq1VNVlVkxMTE4tdnSSpz4Kulqmqo8B9wDuB5UlmLqVcAxxo0weAtQCt/w3A4ZFUK0kayjBXy0wkWd6mTwfeDTxOL+SvbcM2AXe26e1tntZ/b1XVKIuWJM1vmA8xnQNsS7KM3h+D26vqj5LsBm5L8h+BLwO3tPG3AL+bZAr4FrBxDHVLkuYxMNyrahfwjlna99I7/35s+/eA942kOknScfH2Azoxnvj8nF0rn3r5x93PfGHPuKuROsfbD0hSBxnuktRBhrskdZDhLkkdZLhLUgcZ7pLUQYa7JHWQ4S5JHWS4S1IHGe6S1EGGuyR1kOEuSR1kuEtSBxnuktRBhrskdZDhLkkdNMx3qK5Ncl+S3UkeS/Krrf2sJHcn2dOeV7T2JPlkkqkku5JcNO4fQpL0csMcub8E/OuqugC4BLghyQXAjcCOqloP7GjzAFcB69tjC3DzyKuWJM1rYLhX1dNV9XCb/jbwOLAa2ABsa8O2Ade06Q3ArdXzALA8yTkjr1ySNKcFnXNPso7el2U/CKyqqqdb1zPAqja9GtjXt9j+1nbsurYk2Zlk5/T09ALLliTNZ+hwT/JTwB8Av1ZVz/X3VVUBtZANV9XWqpqsqsmJiYmFLCpJGmCocE/yk/SC/feq6g9b88GZ0y3t+VBrPwCs7Vt8TWuTJC2RUwYNSBLgFuDxqvqvfV3bgU3Ax9rznX3tv5zkNuCngWf7Tt9IepU58/CugWNWPv/C3J3Llg+/sfOvGn6s5jUw3IF3Af8U+GqSR1rbv6MX6rcn2Qw8CVzX+u4CrgamgOeB60dasSRpoIHhXlX/D8gc3ZfPMr6AGxZZlyRpEfyEqiR1kOEuSR1kuEtSBxnuktRBhrskdZDhLkkdZLhLUgcZ7pLUQYa7JHWQ4S5JHTTMvWWkV7h/3/2LW8GR3XN27X3hu4tbtySP3CWpiwx3SeogT8tIWrSHX9gzZ9/RI2cMv6J9p8/bfenaS4df12ucR+6S1EGGuyR1kOEuSR00MNyT/HaSQ0ke7Ws7K8ndSfa05xWtPUk+mWQqya4kF42zeEnS7IY5cv80cOUxbTcCO6pqPbCjzQNcBaxvjy3AzaMpU5K0EMN8h+r/TbLumOYNwKVtehtwP/Ch1n5r+x7VB5IsT3JOVT09qoK1tO7ZfXDW9kePHl3Ues887AeVpHE63nPuq/oC+xlgVZteDezrG7e/tb1Cki1JdibZOT09fZxlSJJms+g3VNtReh3HclurarKqJicmJhZbhiSpz/GG+8Ek5wC050Ot/QCwtm/cmtYmSVpCxxvu24FNbXoTcGdf+wfaVTOXAM96vl2Slt7AN1STfIbem6crk+wHPgp8DLg9yWbgSeC6Nvwu4GpgCngeuH4MNUuSBhjmapn3z9F1+SxjC7hhsUVJ6o6908NfGfXcD+e/Cuulb89+9daxrrhg1eBBHecnVCWpgwx3Seogw12SOshwl6QO8ss69GNPfP4VTSufmv0NrjPn+XIGSSeeR+6S1EGGuyR1kKdlJJ00zjy8a97+lc+/MNyKli0fPOb8q4Zb16uUR+6S1EGGuyR1kOEuSR1kuEtSBxnuktRBhrskdZDhLkkd5HXuryL37B58L+tHj37xuNc/6Bpj6UR7eMjbXjz8tcFjnjv84px9b1v+zmFLepmT6T7yHrlLUgeN5cg9yZXAJ4BlwKeq6mPj2I4kHa/5/qc69CdhjzXXJ2NPwKdhRx7uSZYB/x14N7Af+FKS7VW1e9Tbek3ou1PjXHdo7OfdGiXBeI7cLwamqmovQJLbgA3AWMJ9mPPQo7byqXsXvMyw5wolnXjH++/16JEzZu/Yd/qcy1y69tLj2tYg4wj31cC+vvn9wE8fOyjJFmBLm/1OkifGUMvxWAl880QXMQvrWriTtTbrWriTtbYTXdeb5uo4YVfLVNVWYOuJ2v5ckuysqskTXcexrGvhTtbarGvhTtbaTta6YDxXyxwA1vbNr2ltkqQlMo5w/xKwPsl5SU4FNgLbx7AdSdIcRn5apqpeSvLLwBfoXQr521X12Ki3M0Yn3amixroW7mStzboW7mSt7WSti1TVia5BkjRifkJVkjrIcJekDnpNhnuS9yV5LMkPk8x5GVOSK5M8kWQqyY197eclebC1f7a9cTyKus5KcneSPe15xSxjfjbJI32P7yW5pvV9Osk3+vouXKq62rgf9G17e1/7WPbXsLUluTDJF9vvfFeSf9TXN9J9Ntdrpq//tLYPpto+WdfX9+HW/kSSn1tMHcdR1weT7G77Z0eSN/X1zfp7XaK6fiHJdN/2/3lf36b2e9+TZNMo6xqytpv66vpakqN9fWPbZ0OrqtfcA/gbwPnA/cDkHGOWAV8H3gycCnwFuKD13Q5sbNO/BfzSiOr6z8CNbfpG4OMDxp8FfAt4fZv/NHDtGPbXUHUB35mjfSz7a9jagL8OrG/TbwSeBpaPep/N95rpG/MvgN9q0xuBz7bpC9r404Dz2nqWLWFdP9v3Ovqlmbrm+70uUV2/APzmLMueBextzyva9IqlrO2Y8b9C7+KRse6zhTxek0fuVfV4VQ36ROyPbqNQVX8J3AZsSBLgMuCONm4bcM2IStvQ1jfseq8FPl9Vz49o+3NZaF0/Mub9NVRtVfW1qtrTpp8CDgETI6xhxqyvmXnqvQO4vO2jDcBtVfViVX0DmGrrW5K6quq+vtfRA/Q+nzJuw+yvufwccHdVfauqjgB3A1eewNreD3xmhNtftNdkuA9pttsorAbOBo5W1UvHtI/Cqqp6uk0/Awy6OfRGXvmC+o32X+ubkpy2xHW9LsnOJA/MnCpivPtrIbUBkORiekdiX+9rHtU+m+s1M+uYtk+epbePhll2nHX12wx8vm9+tt/rUtb1D9vv544kMx+QHOf+WtD62yms84D+m06Na58NrbNf1pHkHuCvzdL1kaq6c6nrmTFfXf0zVVVJ5rxONck5wN+k93mCGR+mF3Cn0rv+9kPAf1jCut5UVQeSvBm4N8lX6YXXoox4n/0usKmqftiaj3ufdVGSnwcmgZ/pa37F77Wqvj77GkbufwOfqaoXk/wivf/1XLZE2x7WRuCOqvpBX9uJ3GdAh8O9qq5Y5Crmuo3CYWB5klPakdeCbq8wX11JDiY5p6qebkF0aJ5VXQd8rqq+37fumSPYF5P8DvBvlrKuqjrQnvcmuR94B/AHLGJ/jaq2JGcC/4feH/cH+tZ93PtsFsPcemNmzP4kpwBvoPeaGudtO4Zad5Ir6P3B/Jmq+tHXFM3xex1FUA2sq6oO981+it57LDPLXnrMsvePoKaha+uzEbihv2GM+2xonpaZ26y3UajeuyX30TvfDbAJGNX/BLa39Q2z3lec42vhNnOe+xrg0aWqK8mKmVMaSVYC7wJ2j3l/DVvbqcDngFur6o5j+ka5z4a59UZ/vdcC97Z9tB3Y2K6mOQ9YD/zpImpZUF1J3gH8D+C9VXWor33W3+sS1nVO3+x7gcfb9BeA97T6VgDv4eX/ix17ba2+t9J7Q/eLfW3j3GfDO9Hv6J6IB/AP6J1DexE4CHyhtb8RuKtv3NXA1+j9xf1IX/ub6f3DmwJ+HzhtRHWdDewA9gD3AGe19kl632g1M24dvaOInzhm+XuBr9ILqP8J/NRS1QX87bbtr7TnzePeXwuo7eeB7wOP9D0uHMc+m+01Q+80z3vb9OvaPphq++TNfct+pC33BHDViF/zg+q6p/1bmNk/2wf9Xpeorv8EPNa2fx/w1r5l/1nbj1PA9aOsa5ja2vyvAx87Zrmx7rNhH95+QJI6yNMyktRBhrskdZDhLkkdZLhLUgcZ7pLUQYa7JHWQ4S5JHfT/AUJJ9GYMeOhrAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# trained 20 epochs\n",
    "for scale in scales:\n",
    "    plt.figure()\n",
    "    for i in range(repetitions):\n",
    "        plt.hist(results[scale][i].flatten(), alpha = 0.3)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Weights become smaller as scale grows."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "notify_time": "5",
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "position": {
    "height": "407px",
    "left": "1561px",
    "right": "20px",
    "top": "120px",
    "width": "339px"
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}