Figure18.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Figure 18"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Import necessary libraries\n",
    "from __future__ import division, print_function, unicode_literals\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "from mpl_toolkits.mplot3d import Axes3D\n",
    "\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "from functools import partial\n",
    "import glob, os\n",
    "import pickle\n",
    "import copy\n",
    "from PDEsolvers import KS_Periodic"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# User inputs\n",
    "data_train_len = 20  # Number of training data files\n",
    "data_name = 'KS_Eqn_exp5'  # Prefix of data files\n",
    "folder_name = 'KS_exp5h/'  # Folder to be created for saved output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Find the model with lowest validation error\n",
    "import math\n",
    "\n",
    "def BestErrorFile(folder_name):\n",
    "\n",
    "    filename_best = ''\n",
    "    best_error = 10**6\n",
    "\n",
    "    current_dir = os.getcwd()\n",
    "    os.chdir(\"./\" + folder_name)\n",
    "    for file in glob.glob(\"*.pkl\"):\n",
    "        with open(file, 'rb') as f:\n",
    "            params = pickle.load(f)\n",
    "            if not math.isnan(params['minRegTest']):\n",
    "                if params['minRegTest'] < best_error:\n",
    "                    filename_best = file\n",
    "                    best_error = params['minRegTest']\n",
    "    os.chdir(current_dir)\n",
    "\n",
    "    return filename_best"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from ./KS_exp5f/KS_Eqn_exp5_2019_10_16_13_17_10_923832_model.ckpt\n",
      "0.0047629476\n",
      "INFO:tensorflow:Restoring parameters from ./KS_exp6i/KS_Eqn_exp6_2019_10_19_20_18_32_846738_model.ckpt\n",
      "0.024210492\n",
      "INFO:tensorflow:Restoring parameters from ./KS_exp7g/KS_Eqn_exp7_2019_10_21_07_37_53_340646_model.ckpt\n",
      "0.14672409\n",
      "INFO:tensorflow:Restoring parameters from ./KS_exp4f/KS_Eqn_exp4_2019_10_23_18_52_14_520335_model.ckpt\n",
      "0.4687274\n"
     ]
    }
   ],
   "source": [
    "import helperfns\n",
    "\n",
    "# Best full width networks\n",
    "folder_name_list = ['KS_exp5f/','KS_exp6i/','KS_exp7g/','KS_exp4f/']\n",
    "\n",
    "full_width_predloss = np.zeros(4)\n",
    "for ind in range(4):\n",
    "    folder_name = folder_name_list[ind]\n",
    "    pkl_file = folder_name + BestErrorFile(folder_name)\n",
    "\n",
    "    with open(pkl_file, 'rb') as f:\n",
    "         params = pickle.load(f)\n",
    "    \n",
    "    data_name = params['data_name'] \n",
    "    data_test1 = np.load(('./data/%s_test1_x.npy' % data_name))\n",
    "    data_test2 = np.load(('./data/%s_test2_x.npy' % data_name))\n",
    "    data_test3 = np.load(('./data/%s_test3_x.npy' % data_name))\n",
    "    data_test4 = np.load(('./data/%s_test4_x.npy' % data_name))\n",
    "    data_test5 = np.load(('./data/%s_test5_x.npy' % data_name))\n",
    "    data_test_all = np.vstack([data_test1,data_test2,data_test3,data_test4,data_test5])\n",
    "    \n",
    "    max_shifts_to_stack = helperfns.num_shifts_in_stack(params)\n",
    "    data_test_all_tensor = helperfns.stack_data(data_test_all, max_shifts_to_stack, params['val_len_time'])\n",
    "    \n",
    "    tf.reset_default_graph()\n",
    "\n",
    "    with tf.Session() as sess:\n",
    "        saver = tf.train.import_meta_graph(params['model_path']+'.meta')\n",
    "        saver.restore(sess, params['model_path'])\n",
    "    \n",
    "        pred_loss = sess.run(\"loss2:0\", feed_dict={\"x:0\": data_test_all_tensor})\n",
    "        print(pred_loss)\n",
    "        full_width_predloss[ind] = pred_loss\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from ./KS_exp5h/KS_Eqn_exp5_2019_10_18_07_18_03_991009_model.ckpt\n",
      "0.008977019\n",
      "INFO:tensorflow:Restoring parameters from ./KS_exp6j/KS_Eqn_exp6_2019_10_19_20_13_17_886212_model.ckpt\n",
      "0.022841832\n",
      "INFO:tensorflow:Restoring parameters from ./KS_exp7h/KS_Eqn_exp7_2019_10_20_05_52_19_528113_model.ckpt\n",
      "0.13391751\n",
      "INFO:tensorflow:Restoring parameters from ./KS_exp4e/KS_Eqn_exp4_2019_10_22_18_18_45_259709_model.ckpt\n",
      "0.46454412\n"
     ]
    }
   ],
   "source": [
    "# Best reduced width networks\n",
    "folder_name_list = ['KS_exp5h/','KS_exp6j/','KS_exp7h/','KS_exp4e/']\n",
    "\n",
    "reduced_width_predloss = np.zeros(4)\n",
    "for ind in range(4):\n",
    "    folder_name = folder_name_list[ind]\n",
    "    pkl_file = folder_name + BestErrorFile(folder_name)\n",
    "\n",
    "    with open(pkl_file, 'rb') as f:\n",
    "         params = pickle.load(f)\n",
    "            \n",
    "    data_name = params['data_name']  \n",
    "    data_test1 = np.load(('./data/%s_test1_x.npy' % data_name))\n",
    "    data_test2 = np.load(('./data/%s_test2_x.npy' % data_name))\n",
    "    data_test3 = np.load(('./data/%s_test3_x.npy' % data_name))\n",
    "    data_test4 = np.load(('./data/%s_test4_x.npy' % data_name))\n",
    "    data_test5 = np.load(('./data/%s_test5_x.npy' % data_name))\n",
    "    data_test_all = np.vstack([data_test1,data_test2,data_test3,data_test4,data_test5])\n",
    "    \n",
    "    max_shifts_to_stack = helperfns.num_shifts_in_stack(params)\n",
    "    data_test_all_tensor = helperfns.stack_data(data_test_all, max_shifts_to_stack, params['val_len_time'])    \n",
    "    \n",
    "    tf.reset_default_graph()\n",
    "\n",
    "    with tf.Session() as sess:\n",
    "        saver = tf.train.import_meta_graph(params['model_path']+'.meta')\n",
    "        saver.restore(sess, params['model_path'])\n",
    "    \n",
    "        pred_loss = sess.run(\"loss2:0\", feed_dict={\"x:0\": data_test_all_tensor})\n",
    "        print(pred_loss)\n",
    "        reduced_width_predloss[ind] = pred_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWAAAADoCAYAAADR9T5NAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvIxREBQAAIABJREFUeJzt3X10VNW9//H3NyQQohAUrU/YgqBcFRISYiVgKHhBRUFNG4qKDxRbimi5KrbaFbt6769yuxSpD7RXS3n0oRZNLYJKS7VYUsVC4i+ioBSw+BPkCoLBQKAkYf/+OJNhJiSTSTIzh0w+r7XOOnP2efrOXsOXk33O2ducc4iISOKl+B2AiEhHpQQsIuITJWAREZ8oAYuI+EQJWETEJ0rAIiI+UQIWEfGJErCIiE+UgEVEfJLqdwDxcMopp7jevXv7HYaIdFDl5eWfO+dObW67pEzAvXv3pqyszO8wRKSDMrOPo9lOTRAiIj5RAhYRacLWrTBtGnTvDikp3nzaNK88FpSARUQasWIFZGXBvHlgVfv4vSvEqvYxb55XvmJF28+RlG3AjampqWH79u0cOnTI71AkiaWnp9OrVy/S0tL8DkXaYOtWKCqC6mpveQLLKGQpv2c5z9bcSE2Nt379eujbt/XnSaoEbGbjgHH9+vU7Zt327dvp1q0bvXv3xswSH5wkPecce/bsYfv27fTp08fvcKQNZs+Gmpqjy5NZEJw/y42At/6RR+CXv2z9eZKqCcI5t9w5NyUzM/OYdYcOHaJnz55KvhI3ZkbPnj31V1YSGD93FIdrDIc3DeUtAIbxZrDscI3xrSdHtek8SZWAm6PkK/Gm31hy+FldMQfICC534XDYHOAAGfzsyP1tOk+HSsDRivedTxE5vpV1G8lYXg5LwqEOkMFVvEJ5txFtOo8ScAOhdz6rqsA5bx6LO5+dOnVi0KBBwWnbtm0Rt+/duzeff/45ACeeeGKrznnllVdSWVl5TPl//ud/8vDDDwOwaNEiPv3000bP25Q33ngDM2P58uXBsrFjx/LGG29E3K/huWJl0qRJlJSUxPy40jHdeCO8mTaSCSzhIOlh6w6SzgSW8FbaCG66qW3nUQIOEXrnM7QBHrzl6mpvfWuvhLt27UpFRUVwSsTr0q+++io9evSIuE1rk2KvXr2YOXNmi/aJRwKura2N6fFEZsyAtDToQSW1pFJLCtV0pZYUakmlB5WkpcFdd7XtPErAIRre+WxM/Z3PWFm0aBF33HFHcDmaq8h6s2bN4vHHHwfgrrvu4tJLLwXgL3/5CxMnTgTCr2ZnzpzJeeedxyWXXMKmTZsAKCkpoaysjIkTJzJo0CAOHjwIwJw5c8jNzWXgwIF8+OGHjZ4/OzubzMxM/vznPx+zrry8nG984xsMHjyYyy+/nJ07dx5zrtLSUr75zW8C8NJLL9G1a1cOHz7MoUOHOOeccwCoqKhgyJAhZGVlUVhYyBdffAHAiBEjuPPOO8nLy+Oxxx4LO/dPfvITJk2aRF1dXVT1KNJQ375QUgJTUuaTQTXryeYaXmI92WRQzXdTFlBS0rZH0EAJOMwzz0SXgJ9+unXHP3jwYLD5obCwsHUHCVFQUEBpaSkAZWVl7N+/n5qaGkpLSxk+fHjYtuXl5fzud7+joqKCV199lXXr1gFQVFREXl4ezz77LBUVFXTt2hWAU045hXfeeYfbbrst2FTRmOLiYh544IGwspqaGn7wgx9QUlJCeXk5kydPpri4+Jhz5efnU1FRAUBpaSkDBgxg3bp1/P3vf+fiiy8G4Oabb+bBBx9k/fr1DBw4kP/6r/8Knufw4cOUlZUxY8aMYNkPf/hDdu/ezcKFC+nUqVNrq1aEMWNg8KWZLB06i0u7lfGXlNH8e7d1vDT0IS66tDtjxrT9HB3mOeBo7N8f2+0aqm+CiJXBgwdTXl7Ol19+SZcuXcjNzaWsrIzS0tLglXG90tJSCgsLycjwbipcffXVEY9df2U6ePBgXnzxxSa3q0/0f/vb34JlmzZt4v3332f06NEA1NXVccYZZxyzb2pqKn379uWDDz5g7dq13H333axevZq6ujoKCgrYt28flZWVfOMb3wDglltuYfz48cH9J0yYEHa8n/3sZ1x88cXMnTs34ncTidYJf17Kt4BvBUs6ATMCU9slVQJ2zi0Hlufl5X2vNfufeKJ3wy2a7WIlNTWVI0eOBJdb8gxpWloaffr0YdGiRQwdOpSsrCxWrVrFli1bOP/889sUV5cuXQDvxmFzbaz1V8Gpqd7PyTnHhRdeyJo1a5o9z/Dhw1mxYgVpaWmMGjUq2HQwa9asZvc94YQTwpYvuugiysvL2bt3LyeffHKz+4v4TU0QIW680Wt4jyQtjTbf+QzVu3dvKioqOHLkCJ988glr165t0f4FBQU8/PDDDB8+nIKCAp588klycnKOeR51+PDhLF26lIMHD1JVVRX29EK3bt2oiuZ/niZcdtllfPHFF6xfvx6A/v37s3v37mACrqmpYcOGDY2eq6CggEcffZT8/HxOPfVU9uzZw6ZNmxgwYACZmZmcdNJJwWaWp59+Ong13JgrrriC++67j6uuuqpN30ckUZLqCritZsyAxYsjtwPH4s5nqGHDhtGnTx8uuOACzj//fHJzc1u0f0FBATNnziQ/P58TTjiB9PR0CgoKjtkuNzeXCRMmkJ2dzVe+8hUuuuii4LpJkyYxdepUunbtGtVVa2OKi4u55pprAOjcuTMlJSVMnz6dffv2UVtby5133smFF154zLkuvvhiPvvss2BTRlZWFv/7v/8b/A9k8eLFTJ06lerqas455xwWLlwYMY7x48dTVVXF1Vdfzauvvhps0xY5Hplzzu8YYi4vL8817JD9gw8+iOrP8hUrvEfNamrCE3FamjeVlBCTxndJXtH+1iR5mVm5cy6vue3UBNHAmDFeD0dTpoS/CTdlileu5CsisaImiEb07ev1cNSWXo5ERJqjK2AREZ8oAYuI+EQJWETEJ0mVgM1snJnN3bdvX2wOuG8fFBZ6cxGRGEuqBBxpRIxWWbYMli6FkJcW2qK+O8oBAwYwbty4RruJjCS0C8l42LZtGwMGDGjVvkOHDm20PLSbyEcffZTq+kG2iK6LzUWLFpGSkhJ8yQNgwIABzXbl2fBcsTJixAgaPuIo0lpJlYBjbsGC8Hkb1fcF8f7773PyySfzq1/9KibHPR689dZbzW7T2qTYmm4v45GA1buaxJoScKhRo8Ds6FSfVN58M7x8VNvGgQLIz89nx44dweVZs2Zx0UUXkZWVxU9/+tNgeWNdSEL4ldjnn38e7Fu4rq6Oe+65hwEDBpCVlcWcOXOAxruHrC/Pzs4mOzu7yf8Qbr/9dpYtWwZAYWEhkydPBmDBggUUFxcDR69mnXPccccd9O/fn1GjRrFr1y4AHn/8cT799FNGjhzJyJEjg8cuLi4mOzubIUOG8NlnnzV6/rFjx7Jhw4aw719v5cqV5Ofnk5uby/jx49m/f/8x53rhhRe4++67AXjssceCXV1+9NFHDBs2DIDXX3+dnJwcBg4cyOTJk/nXv/4FeK+K33vvveTm5vLCCy8Ez3vkyBEmTZrE/fe3bUga6diUgEMVF0NGyBAkhw+Hz8Fb38Z/dHV1dbz++uvBHslWrlzJ5s2bWbt2LRUVFZSXl7N69eomu5CMZO7cuWzbto2KigrWr1/PxIkTm+weEuA73/kOc+bM4d13323ymKHdXu7YsYONGzcCNNrt5R/+8Ac2bdrExo0beeqpp4JXxtOnT+fMM89k1apVrFq1CoADBw4wZMgQ3n33XYYPH85vfvObRs+fkpLCj370I/77v/87rPzzzz/ngQce4LXXXuOdd94hLy+PX/ziF8ecKzT+0tJSevbsyY4dO4LxHzp0iEmTJrFkyRLee+89amtreeKJJ4Ln6dmzJ++88w7XXXcd4HUAP3HiRM4999xjuuIUaQkl4FAjR8LLL4cn4VAZGfDKKzBiRKsOX98f8Omnn85nn30W7K5x5cqVrFy5kpycHHJzc/nwww/ZvHlzWBeS3bt3b7YLSYDXXnuN73//+8GeyU4++eSw7iEHDRrEAw88wPbt26msrKSysjKYRG9qopeh+gS2ceNGLrjgAk477TR27tzJmjVrjmn7Xb16Nddffz2dOnXizDPPDHYS35jOnTszduxYwOv2MlK77g033MDbb7/NP//5z2DZ22+/zcaNGxk2bBiDBg1i8eLFfPzxx8fse/rpp7N//36qqqr45JNPuOGGG1i9ejWlpaUUFBSwadMm+vTpw3nnnQd43V6uXr06uH/Dbi+///3vM2DAgOB/YiKtpQTc0MiRsGQJpIePA0V6ulfeyuQLR9uAP/74Y5xzwT/5nXP8+Mc/Dg5VtGXLFm699daIxwrtxrK5Lizru4esP/57773HypUro477rLPOorKykj/+8Y/BXteef/55TjzxRLp16xb1cRpKS0sLdrrTXLeXqampzJgxgwcffDDse40ePTr4vTZu3Mj8+fMb3X/o0KEsXLiQ/v37B/9DWbNmTbAJIpKG3V4OHTqUVatWafh5aTMl4MZUVkJqqtcRRNeu3jw11SuPgYyMDB5//HFmz55NbW0tl19+OQsWLGB/oKf3HTt2sGvXrohdSPbu3Zvy8nKAsMEoR48eza9//etgMtu7d2+T3UP26NGDHj16BDtTf/bZZ5uMeciQITz66KPBBPzwww832uva8OHDWbJkCXV1dezcuTPY3ABt7/Zy0qRJvPbaa+zevTsY05tvvsmWLVsAr0njH//4R6PnCu22Mycnh1WrVtGlSxcyMzPp378/27ZtCx6nuW4vb731Vq688kq+/e1vazw6aRMl4MbMn++NwJmdDS+95M2rq2P2NARATk4OWVlZPPfcc1x22WXccMMN5OfnM3DgQIqKiqiqqgrrQnLMmDFhXUjec889PPHEE+Tk5ISNYPzd736Xr371q2RlZZGdnc1vf/vbYPeQ9957L9nZ2QwaNCjYNrtw4UJuv/12Bg0aRKSe8QoKCqitraVfv37k5uayd+/eRhNwYWEh5557LhdccAE333wz+fn5wXVTpkzhiiuuCLsJ1xKdO3dm+vTpwRt7p556KosWLeL6668nKyuL/Pz84Ph1Dc9VUFDAJ598wvDhw+nUqRNnn302l1xyCQDp6eksXLiQ8ePHM3DgQFJSUpg6dWrEWO6++25ycnK46aabwjrUF2kJdUfZmGuvheHD4c47vavfujp49FEoLfWeCxaJQN1RSrTdUao3tMY0TLKdOnm9tc+IzThQIiKgJggREd90qAScjM0tcnzRb0xaosMk4PT0dPbs2aN/IBI3zjn27NlDesNHGEWa0GHagHv16sX27duDjzCJxEN6ejq9evXyOwxpJ5IqAZvZOGBcv379jlmXlpZGnz59Eh+UiEgTkqoJIubdUYqIxFFSJWARkfZECVhExCdKwCIiPlECFhHxiRKwSBLZuhWmTYPu3b1uTLp395a3bvU7MmmMErBIklixArKyYN48qKoC57z5vHle+YoVfkcoDSkBiySBrVuhqMjrNbWmJnxdTY1XXlSkK+HjjRKwSBKYPTs88XZnHy9SSHf2BctqauCRR3wITpqkBCySBJ55JjwBX80yClnKOI6OolJTA08/7UNw0iQlYJEkEBjNKmgyC8LmTW0n/lICFkkCr6eMwmHBaSjekFPDeDOs/DUb5XOkEkoJWCQJvHtVMQfICC534XDYHOAAGawfd3/CY5OmKQGLJIFxvxhJUZeXw5JwqANk8K0urzD24RGJDUwiUgIWSQJ9+8L0P4zk5i5LOEh4h/AHSefmLkv4jz+MoG9fnwKURjWbgM1svJl1C3y+38xeNLPc+IcmIi0xZgz8z8xKLC2VWlKopiu1pGBpqfzPzErGjPE7QmkomivgnzjnqszsEmAUMB94Ir5hiUhrnPbyfNLrqknNySZj5Uuk5mSTXlfNaa8saH5nSbhoEnBdYH4VMNc59wrQOX4hiUirZWbCrFlQVgajR8O6dfDQQ16nEHLcseYGqTSzl4EdwGggFzgIrHXOZcc/vNbJy8tzZWVlfochIh2UmZU75/Ka2y6aK+BvA38CLnfOVQInAz9sY3wiIh1eNAn4DOAV59xmMxsBjAfWxjWqEGZ2jpnNN7OSRJ1TRCQRoknAvwfqzKwfMBc4G/htNAc3swVmtsvM3m9QfoWZbTKzLWZ2X6RjOOc+cs7dGs35RETak2iGpT/inKs1s28Cc5xzc8zs/0Z5/EXAL4Gn6gvMrBPwK7w25e3AOjNbBnQCft5g/8nOuV1RnktEpF2JJgHXmNn1wM3AuEBZWjQHd86tNrPeDYq/Dmxxzn0EYGa/A65xzv0cGBvNcUVEkkE0TRDfAfKBmc65f5pZH6AtndqdBXwSsrw9UNYoM+tpZk8COWb24wjbTTGzMjMr2717dxvCExFJjGavgJ1zG83sHuA8MxsAbHLOPRj/0ILn3wNMjWK7uXht1OTl5UV+tk5E5DjQbAIOPPmwGNgGGHC2md3inFvdynPuwLuRV69XoExEpEOJpg14NnCZc24TgJmdBzwHDG7lOdcB5waaMnYA1wE3tPJYIiLtVjRtwGn1yRfAOfcPorwJZ2bPAWuA/ma23cxudc7VAnfgvdzxAfC8c25Dy0Nv9HzjzGzuvn37mt9YRMRn0byKvAA4AjwTKJoIdHLOTY5zbK2mV5FFxE/RvoocTRPEbcDtwPTAcinec7wiItIG0TwF8S/gF4EJADNbAkyIY1wiIkmvtSNi5Mc0ihhRG7CItCdJNSSRc265c25KZmam36GIiDSrySaICMMOGVE+BSEiIk2L1AY8O8K6D2MdiIhIR9NkAnbOjUxkICIiHU1StQHrJpyItCdJlYB1E05E2pOkSsAiIu1JNG/CYWZnAV8L3b4NvaGJiAjRdUf5IN5bbxuBukCxA5SARUTaIJor4GuB/oFXko9rZjYOGNevXz+/QxERaVY0bcAf0U5evNBNOBFpT6K5Aq4GKszsdSB4Feycm970LiIi0pxoEvCywCQiIjEUTXeUi82sM3BeoGiTc64mvmGJiCQ/PwblFBER/BmUU0REiPOgnImmviBEpD2JJgGXmdk8MxsRmH4DHJcjXuoxNBFpT1o7KOf/xC0iEZEOolWDcoqISNtFGpLoeefct83sPby+H8I457LiGpmISJKLdAX8H4H52EQEIiLS0TR5E845tzPwcZpz7uPQCZiWmPBERJJXNE9BjG6kbEysAxER6WgitQHfhnel29fM1oes6ga8Fe/AWkPdUYpIe2LOHXN/zVthlgmcBPwcuC9kVZVzbm8CYmu1vLw8V1Z2XD6qLCIdgJmVO+fymtsuUhvwPufcNuAxYG9I+2+tmV0cu1BFRDqmaNqAnwD2hyzvD5SJiEgbRJOAzYW0UzjnjhDlYJ4iItK0qIYkMrPpZpYWmP4Db5giERFpg2gS8FRgKLAD2A5cDEyJZ1AiIh1BNH1B7AKuS0AsIiIdSqTngH/knHvIzObQeF8QGpRTRKQNIl0BfxCY64FaEZE4aDIBO+eWB+aLExdO2+hNOBFpTyI1QSynkaaHes65q+MSURsE/tNYnpeX9z2/YxERaU6kJoiHA/NvAqcDzwSWrwc+i2dQIiIdQaRXkf/qnPsrMMw5NyEw3tpy59wNQEHiQpRktnUrTJsG3btDSoo3nzbNKxdJdtE8B3yCmZ1Tv2BmfYAT4heSdBQrVkBWFsybB1VV4Jw3nzfPK1+xwu8IReIrmgR8F/CGmb1hZn8FVgF3xjcsSXZbt0JREVRXQ00NdGcfL1JId/ZRU+OVFxXpSliSW7MJ2Dn3R+BcvCGKpgP9nXN/indgktxmz/YSb72rWUYhSxnH8mBZTQ088ogPwYkkSLMJ2MwygB8Cdzjn3gW+amYaJ07a5JlnwhPwZBaEzcFb//TTiY5MJHGiaYJYCBwG8gPLO4AH4haRdAgvVo3CYcFpaGCQlWG8GVb++y9H+RypSPxEk4D7OuceAmoAnHPVgMU1Kkl6j3Qt5gAZweUuHA6bAxwgg0cy7k94bCKJEk0CPmxmXQm8lGFmfYF/xTUqSXpfmzSSazu9HJaEQx0gg2s6vUKf74xIbGAiCRRNAv4p8EfgbDN7Fngd+FFco5KkN2MGvNVlJBNYwkHSw9YdJJ0JLGFNlxHcdZdPAYokQMQEbGYGfIj3Ntwk4Dkgzzn3Rtwjk6TWty+UlMBpnSupJZVaUqimK7WkUEsqp3WupKTE204kWUVMwIGhiF51zu1xzr3inHvZOfd5gmKTJDdmDMzJmc8JVLMhJZtCe4kNKdmcQDVzchcwZozfEYrEVzRNEO+Y2UVxj0Q6pIzTM0mZPYvsmjL+dGQ02YfXkfLwQ2Sc1t3v0ETizkLG22x8A7MP8V7E2AYcwHsCwjnnsuIeXQuFdEf5vc2bN/sdjoh0UGZW7pzLa267aEY3vjwG8SSEuqMUkfYkUn/A6XgDcvYD3gPmO+dqExWYiEiyi9QGvBjIw0u+Y4DZCYlIRKSDiNQEcYFzbiCAmc0H1iYmJBGRjiHSFXCwqxQ1PYiIxF6kK+BsM/sy8NmAroHl+qcg9JyQiEgbRBoVuVMiAxER6WiieRFDRETiQAlYRMQnSsAiIj5RAhYR8YkSsIiIT5SARUR8ogQsIuITJWAREZ8oAYuI+EQJWETEJ0rAIiI+UQIWEfGJErCIiE+iGRPOV2Z2LXAV0B1vWKSVPockIhITcb0CNrMFZrbLzN5vUH6FmW0ysy1mdl+kYzjnljrnvoc3Pt2EeMYrIpJI8b4CXgT8EniqvsDMOgG/AkYD24F1ZrYM6AT8vMH+k51zuwKf7w/sJyKSFOJ6BeycWw3sbVD8dWCLc+4j59xh4HfANc6595xzYxtMu8zzILDCOfdOPONtsX37oLDQm4uItJAfN+HOAj4JWd4eKGvKD4BRQJGZTW1qIzObYmZlZla2e/fu2ETanGXLYOlSWL48MecTkaRy3D8F4Zx73Dk32Dk31Tn3ZITt5jrn8pxzeaeeempigluwIHwuItICfiTgHcDZIcu9AmXHv1GjwOzo9NZbXvmbb4aXjxrlb5wi0i74kYDXAeeaWR8z6wxcByzzIY6WKy6GjIyjy4cPh8/BW3///YmNS0TapXg/hvYcsAbob2bbzexW51wtcAfwJ+AD4Hnn3IYYnW+cmc3dF6+bYiNHwssvhyfhUBkZ8MorMGJEfM4vIknFnHN+xxBzeXl5rqysLH4nePllGD8eDh06WpaeDi+8AGPHxu+8ItIumFm5cy6vue2O+5twx6XKSkhNhZQU6NrVm6emeuUiIlFSAm6N+fOhuhqys+Gll7x5dbWehhCRFkmqBBz3NuB6mZkwaxaUlcHo0bBuHTz0EHTvHt/zikhSURuwiEiMqQ1YROQ4pwQsIuITJWAREZ8kVQJO2E04EZEYSKoE7Jxb7pybkpmZ6XcoIiLNSqoELCLSnigBi4j4RAlYRMQnSZWAdRNORNqTpErAugknIu1JUiVgEZH2RAlYRMQnSsCtsHUrTJvmdX6WkuLNp03zykVEoqUE3EIrVkBWFsybB1VV4Jw3nzfPK1+xwu8IRaS9UAJuga1boajI63u9piZ8XU2NV15UpCthEYlOUiXgeD+GNnv2sYm3oZoaeOSRuJxeRJKMOmRvge7dveaGaLbTo8giHZc6ZI+D/ftju52IdGxKwC1w4omx3U5EOjYl4Ba48UZIS4u8TVoa3HRTYuIRkfZNCbgFZsyILgHfdVdi4hGR9k0JuAX69oWSEsjIODYRp6V55SUl3nYiIs1RAm6hMWNg/XqYMiX8TbgpU7zyMWP8jlBE2oukegzNzMYB4/r16/e9zZs3+x2OiHRQHfIxNHVHKSLtSVJdAdczs93Axwk63SnA5wk6lx8ygUS9VhKvukzkd2hOImP5KvD/4nDcWH6Hth6rtfu3Zr+W/D6/5pw7tbmNkjIBJ5KZlUXzp0Z7ZWZznXNTEnSuuNRlIr9DcxJcn7ujSQKtOG7MvkNbj9Xa/VuzXzx+n0nVBCFxsdzvAGLgePoOiYylMk7HjeV3aOuxWrv/cfGb0BVwGyX7FXAiqS5jS/UZW7oCPj7N9TuAJKK6jC3VZ2zFvD51BSwi4hNdAYuI+EQJOISZXWFmm8xsi5nd18j64Wb2jpnVmllRSPkgM1tjZhvMbL2ZTQhZt8jM/mlmFYFpUKK+j9+iqM+7zWxjoM5eN7OvhayrC6mzZSHlfczs74FjLjGzzon6Pn6Koi4nmdnukDr7bsi6W8xsc2C6JaR8sJm9Fzjm42Zmifo+fjOzBWa2y8zeb2K9BepkS+D3mRuyLnb16ZzT5DXDdAK2AucAnYF3gQsabNMbyAKeAopCys8Dzg18PhPYCfQILC8K3bajTFHW50ggI/D5NmBJyLr9TRz3eeC6wOcngdv8/q7HSV1OAn7ZyL4nAx8F5icFPp8UWLcWGAIYsAIY4/d3TWCdDgdygfebWH9loE4sUEd/j0d96gr4qK8DW5xzHznnDgO/A64J3cA5t805tx440qD8H865zYHPnwK7gJg/f9nORFOfq5xz1YHFt4FekQ4YuKK4FCgJFC0Gro1p1MenZusygsuBPzvn9jrnvgD+DFxhZmcA3Z1zbzsvezxFx6hLAJxzq4G9ETa5BnjKed4GegTqLKb1qQR81FnAJyHL2wNlLWJmX8e7SgkdmnNm4M+YR8ysS9vCbDdaWp+34l011Es3szIze9vM6n/IPYFK51xtlMdMFtHW5bcCv7MSMzu7mX3PCnxu7pgdVaR6i1l9KgHHUOB/waeB7zjn6q+Sfwz8G3AR3p8t9/oU3nHLzG4E8oBZIcVfc94zlzcAj5qZOvmMbDnQ2zmXhXdVttjneCQKSsBH7QDODlnuFSiLipl1B14BigN/sgDgnNsZ+DPmX8BCvD8nO4Ko6tPMRgHFwNWBOgLAObcjMP8IeAPIAfbg/SmYGumYSajZunTO7Qmpv3nA4Gb23UF4k09HqctoRaq3mNWnEvBR64BzA3fZOwPXAcua2QeAwPZ/wGszKmmw7ozA3PDahBq965qEmq1PM8sBfo2XfHeJBoQJAAACG0lEQVSFlJ9U31RjZqcAw4CNgba1VUD9Eyi3AC/F/Zv4L5q6PCNk8Wrgg8DnPwGXBer0JOAy4E/OuZ3Al2Y2JPDbvJmOUZfRWgbcHHgaYgiwL1Bnsa1Pv+9GHk8T3p3Pf+C13xYHyv4PXoIArxlhO3AA72psQ6D8RqAGqAiZBgXW/QV4Dy/xPgOc6Pf3PI7q8zXgs5A6WxYoHxqos3cD81tDjnkO3t3mLcALQBe/v+dxUpc/BzYE6mwV8G8h+04O1NcWvOax+vK8wO9yK/BLAi9mdYQJeA7vaaWawL/pW4GpwNTAegN+Faib94C8eNSn3oQTEfGJmiBERHyiBCwi4hMlYBERnygBi4j4RAlYRMQnSsAiIj5RAhYR8YkSsAhgZteamTOzfwsp6xXat7NIrCkBi3iuB8oC83r/jtdnrEhc6E046fDM7ES810cvA15wzp1nZpfgvctfCVQB33Rex0AiMaMrYBGv8+3XnHPvAvvNbLBz7m94neBc45wbpOQr8aAELOI1Ozwf+Pw8R5sh+gMf+hKRdAhKwNKhmdnJwMXAHwNFzwMTzOxUvC4Ia5vcWaSNlICloysCXnWBzswDTQ07gfOBT/0MTJKfbsJJh2Zmq4Bs4MuQ4p54A39eCGQAU5xzb/kQniQ5JWAREZ+oCUJExCdKwCIiPlECFhHxiRKwiIhPlIBFRHyiBCwi4hMlYBERnygBi4j45P8DdMjl0dGjTEMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 360x239.76 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "dt_array = np.array([0.125, 0.25,0.5,1])\n",
    "\n",
    "plt.figure(figsize=(5,3.33))\n",
    "Full, = plt.plot(dt_array,full_width_predloss,'ob',markersize=10)\n",
    "Reduced, = plt.plot(dt_array,reduced_width_predloss,'*r',markersize=10)\n",
    "plt.xscale('log')\n",
    "plt.yscale('log')\n",
    "plt.rcParams['axes.labelsize'] = 10\n",
    "plt.rcParams['xtick.labelsize'] = 10\n",
    "plt.rcParams['ytick.labelsize'] = 10\n",
    "plt.xticks([0.125,0.25,0.5,1.0,2])\n",
    "ax = plt.gca()\n",
    "ax.get_xaxis().set_major_formatter(matplotlib.ticker.ScalarFormatter())\n",
    "plt.xlim(0.1,1.1)\n",
    "plt.xlabel('$\\Delta t$')\n",
    "plt.ylabel('Prediction Loss')\n",
    "plt.legend([Full,Reduced],['Full width Network','Reduced width Network'],fontsize=10)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.00476295 0.02421049 0.14672409 0.46872741]\n",
      "[0.00897702 0.02284183 0.13391751 0.46454412]\n"
     ]
    }
   ],
   "source": [
    "print(full_width_predloss)\n",
    "print(reduced_width_predloss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tensorflow",
   "language": "python",
   "name": "tensorflow"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}