mnist-inception-results.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "%matplotlib inline\n",
    "pd.set_option('display.max_colwidth', -1)\n",
    "pd.set_option(\"display.max_rows\", None)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Useful Functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_training_log(log, acc=True, loss=True, lr=True, figsize=(10.0, 3.0)):\n",
    "    plt.rcParams['figure.figsize'] = figsize\n",
    "    i = log['val_acc'].idxmax() + 1\n",
    "    \n",
    "    if acc:\n",
    "        plt.axvline(x=i, color='0.5', linestyle='--')\n",
    "        plt.plot(log['epoch'], log['acc'], label='Train')\n",
    "        plt.plot(log['epoch'], log['val_acc'], label='Validation')\n",
    "        plt.minorticks_on()\n",
    "        plt.grid(b=True, axis='x', which='both', color='0.8', linestyle='-')\n",
    "        plt.xlabel('Epoch')\n",
    "        plt.ylabel('Accuracy')\n",
    "        plt.xlim(0, log['epoch'].iloc[-1] + 1)\n",
    "        #plt.ylim(0.9, 1.0)\n",
    "        plt.legend()\n",
    "        plt.tight_layout()\n",
    "        plt.show()\n",
    "        \n",
    "    if loss:\n",
    "        plt.axvline(x=i, color='0.5', linestyle='--')\n",
    "        plt.plot(log['epoch'], log['loss'], label='Train')\n",
    "        plt.plot(log['epoch'], log['val_loss'], label='Validation')\n",
    "        plt.minorticks_on()\n",
    "        plt.grid(b=True, axis='x', which='both', color='0.8', linestyle='-')\n",
    "        plt.xlabel('Epoch')\n",
    "        plt.ylabel('Loss')\n",
    "        plt.xlim(0, log['epoch'].iloc[-1] + 1)\n",
    "        #plt.ylim(0.9, 1.0)\n",
    "        plt.legend()\n",
    "        plt.tight_layout()\n",
    "        plt.show()\n",
    "    \n",
    "    if lr and 'lr' in log.columns:\n",
    "        fig, ax1 = plt.subplots()\n",
    "        plt.axvline(x=i, color='0.5', linestyle='--')\n",
    "\n",
    "        ln1 = ax1.plot(log['epoch'], log['lr'], 'C4o-', label='Learning Rate')   \n",
    "        ax1.set_xticks(log['epoch'], minor=True)\n",
    "        ax1.grid(b=True, axis='x', which='both', color='0.8', linestyle='-')\n",
    "        ax1.set_xlabel('Epoch')\n",
    "        ax1.set_ylabel('Learning Rate')\n",
    "        ax1.set_xlim(0, log['epoch'].iloc[-1] + 1)\n",
    "\n",
    "        ax2 = ax1.twinx()\n",
    "        ln2 = ax2.plot(log['epoch'], log['val_acc'], 'C1o-', \n",
    "                       label='Validation Accuracy [0.98 - 1.0]')        \n",
    "        #ax2.set_ylabel('Val Accuracy')\n",
    "        ax2.set_yticks([])\n",
    "        ax2.set_ylim(0.98, 1.0)        \n",
    "\n",
    "        lns = ln1 + ln2        \n",
    "        plt.legend(lns, [l.get_label() for l in lns])\n",
    "        plt.tight_layout()\n",
    "        plt.show()\n",
    "        \n",
    "        \n",
    "def plot_log_for(results, idx=-1, acc=True, loss=True, lr=True, figsize=(10.0, 3.0)):   \n",
    "    filename = results.iloc[idx]['log']\n",
    "    if type(filename) != str:\n",
    "        print('Log not available')\n",
    "        return\n",
    "    log = pd.read_csv('logs/' + filename)\n",
    "    log['epoch'] += 1\n",
    "    display(results.loc[[idx]].drop(['epochs', 'log'], axis=1))\n",
    "    plot_training_log(log, acc=acc, loss=loss, lr=lr, figsize=figsize)\n",
    "    return idx, log\n",
    "    \n",
    "    \n",
    "def plot_annealers(annealers, epochs=40, figsize=(10.0, 3.0)):\n",
    "    plt.rcParams['figure.figsize'] = figsize\n",
    "    x = np.arange(0, epochs)\n",
    "    for k, v in annealers.items():\n",
    "        plt.plot(x + 1, v(x), label=k)\n",
    "    plt.minorticks_on()\n",
    "    plt.grid(b=True, axis='x', which='both', color='0.8', linestyle='-')\n",
    "    plt.grid(b=True, axis='y', which='major', color='0.8', linestyle='-')\n",
    "    plt.xlabel('Epoch')\n",
    "    plt.ylabel('Learning Rate')\n",
    "    plt.xlim(0, epochs + 1)\n",
    "    plt.legend()\n",
    "    plt.tight_layout()\n",
    "    plt.show()\n",
    "    \n",
    "    \n",
    "def reduce_on_plat(lr, factor, plats, epochs):\n",
    "    y = []\n",
    "    for e in range(epochs):\n",
    "        if e in plats:\n",
    "            lr *= factor\n",
    "        y.append(lr)\n",
    "    return y"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Exploratory Analysis"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "results = pd.read_csv('results.csv')\n",
    "results['notes'].fillna('', inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>val_acc</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>lr</th>\n",
       "      <th>epoch</th>\n",
       "      <th>optimizer</th>\n",
       "      <th>optimizer args</th>\n",
       "      <th>annealer</th>\n",
       "      <th>annealer args</th>\n",
       "      <th>dropout type</th>\n",
       "      <th>dropout rates</th>\n",
       "      <th>notes</th>\n",
       "      <th>batch size</th>\n",
       "      <th>epochs</th>\n",
       "      <th>keras</th>\n",
       "      <th>tf</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.9945</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0413</td>\n",
       "      <td>16</td>\n",
       "      <td>SGD</td>\n",
       "      <td>(lr=0.066, momentum=0.95)</td>\n",
       "      <td>LinearDecay</td>\n",
       "      <td>NaN</td>\n",
       "      <td>spatial</td>\n",
       "      <td>(0,0,.3,.5)</td>\n",
       "      <td></td>\n",
       "      <td>64</td>\n",
       "      <td>40</td>\n",
       "      <td>2.1.5</td>\n",
       "      <td>1.4.1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    val_acc  val_loss      lr  epoch optimizer             optimizer args  \\\n",
       "16  0.9945  NaN        0.0413  16     SGD       (lr=0.066, momentum=0.95)   \n",
       "\n",
       "       annealer annealer args dropout type dropout rates notes  batch size  \\\n",
       "16  LinearDecay  NaN           spatial      (0,0,.3,.5)         64           \n",
       "\n",
       "    epochs  keras     tf  \n",
       "16  40      2.1.5  1.4.1  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rfilter = (results['optimizer'].str.contains('SGD')) & (results['annealer'].str.contains('Linear')) # & (results['log'].notna())\n",
    "sort_by = ['val_acc', 'epoch', 'optimizer args']\n",
    "sort_ascending = [0,1,1]\n",
    "\n",
    "results[rfilter].drop('log', axis=1).sort_values(by=sort_by, ascending=sort_ascending)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>val_acc</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>lr</th>\n",
       "      <th>epoch</th>\n",
       "      <th>optimizer</th>\n",
       "      <th>optimizer args</th>\n",
       "      <th>annealer</th>\n",
       "      <th>annealer args</th>\n",
       "      <th>dropout type</th>\n",
       "      <th>dropout rates</th>\n",
       "      <th>notes</th>\n",
       "      <th>batch size</th>\n",
       "      <th>keras</th>\n",
       "      <th>tf</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>0.995</td>\n",
       "      <td>0.0251</td>\n",
       "      <td>0.015</td>\n",
       "      <td>25</td>\n",
       "      <td>SGD</td>\n",
       "      <td>(lr=0.06, momentum=0.96, nesterov=False)</td>\n",
       "      <td>ReduceLROnPlateau</td>\n",
       "      <td>(monitor='val_loss', factor=0.5, patience=4, min_lr=0.001, verbose=1)</td>\n",
       "      <td>spatial</td>\n",
       "      <td>(0,0,.3,.5)</td>\n",
       "      <td></td>\n",
       "      <td>64</td>\n",
       "      <td>2.2.2</td>\n",
       "      <td>1.10.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    val_acc  val_loss     lr  epoch optimizer  \\\n",
       "44  0.995    0.0251    0.015  25     SGD        \n",
       "\n",
       "                              optimizer args           annealer  \\\n",
       "44  (lr=0.06, momentum=0.96, nesterov=False)  ReduceLROnPlateau   \n",
       "\n",
       "                                                            annealer args  \\\n",
       "44  (monitor='val_loss', factor=0.5, patience=4, min_lr=0.001, verbose=1)   \n",
       "\n",
       "   dropout type dropout rates notes  batch size  keras      tf  \n",
       "44  spatial      (0,0,.3,.5)         64          2.2.2  1.10.0  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x180 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x180 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>val_acc</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>lr</th>\n",
       "      <th>epoch</th>\n",
       "      <th>optimizer</th>\n",
       "      <th>optimizer args</th>\n",
       "      <th>annealer</th>\n",
       "      <th>annealer args</th>\n",
       "      <th>dropout type</th>\n",
       "      <th>dropout rates</th>\n",
       "      <th>notes</th>\n",
       "      <th>batch size</th>\n",
       "      <th>keras</th>\n",
       "      <th>tf</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>0.9948</td>\n",
       "      <td>0.024</td>\n",
       "      <td>0.033</td>\n",
       "      <td>29</td>\n",
       "      <td>SGD</td>\n",
       "      <td>(lr=0.066, momentum=0.96, nesterov=False)</td>\n",
       "      <td>ReduceLROnPlateau</td>\n",
       "      <td>(monitor='val_loss', factor=0.5, patience=3, min_lr=0.0002, verbose=1)</td>\n",
       "      <td>spatial</td>\n",
       "      <td>(0,0,.3,.5)</td>\n",
       "      <td>padding='same' in stem</td>\n",
       "      <td>64</td>\n",
       "      <td>2.1.5</td>\n",
       "      <td>1.4.1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    val_acc  val_loss     lr  epoch optimizer  \\\n",
       "34  0.9948   0.024     0.033  29     SGD        \n",
       "\n",
       "                               optimizer args           annealer  \\\n",
       "34  (lr=0.066, momentum=0.96, nesterov=False)  ReduceLROnPlateau   \n",
       "\n",
       "                                                             annealer args  \\\n",
       "34  (monitor='val_loss', factor=0.5, patience=3, min_lr=0.0002, verbose=1)   \n",
       "\n",
       "   dropout type dropout rates                   notes  batch size  keras  \\\n",
       "34  spatial      (0,0,.3,.5)   padding='same' in stem  64          2.1.5   \n",
       "\n",
       "       tf  \n",
       "34  1.4.1  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x180 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsgAAACsCAYAAABmUVoTAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvIxREBQAAIABJREFUeJzs3Xlc1NX6wPHPYQdRBBVNwSVLcUNccmkxzbSs3Mo0s8XW23brtnfr1+atu1Vm3babpZZZZotLZlYuZbfMHYkEd1PcQFEQWQTm/P44wz7DDDMMw8Dzfr3mBXxnzvd7OHyVhzPPeY7SWiOEEEIIIYQw/LzdASGEEEIIIeoTCZCFEEIIIYQoRwJkIYQQQgghypEAWQghhBBCiHIkQBZCCCGEEKIcCZCFEEIIIYQoRwJkIYQQQgghypEAWQghhBBCiHIkQBZCCCGEEKKcAG93oLb4+fnp0NBQl9tbLBb8/Fz/e0Ha+257X+67tJd7R9o3zvbutC0qKgIgIMD1EMCXx07aN+7/93Jzc7XW2vEJtNYN4hESEqLdsW3btkbdfvPmzV69vjfby9jJveeL19bat8euPrT35fFzp+3s2bP1m2++6XJ7rX177OpDe18eP18fO+C0diKulBQLIYQQQgghymkwKRau2rHuCGsX7yYns4D1UZkMHtuZLgPbNMr2SVE/+3T/a9pexq7+tPe18ZOxc6+9EELUd406QN6x7gir56VSdMYCQE5mAavmpnAyI5f23Vs4bL9/23E2L/+D4iIt7X2svS/3XdrLvVPf2q+elwogQbKPaNOmDZmZmd7uhhD1mjLpGL4vNDRU5+XlVThWWFhIWloa+fn5NtvknChAWxrG9y+ELVpDbmYRO1bnUJQv97rwnPCoYG7++wU1apOSkkK3bt1cvqa77bds2UKfPn28dn132svY+XZ7Xx4/Xx87pVSu1rqJo9c16BnktLQ0mjZtSseOHVFKVXk+/Y9su20josMcnj8rPVfa+2h7X+57TdprrTlxMpPWnU7SxL9l6fGl/9lqt/1Vf+7t8PqNub0v992T7XMyCxy2FUIIX9GgA+T8/Hy7wTGAn78flmKLzePBoY6HRtr7bntf7ntN27cJjeZkViYdupW9fR4eFWwzoAmPCqZDD8dvszfm9r7cd0+3F77hyy+/JCsry61ZPCEaugZfxcJecAwQ3jy4yvNKKcKbO/cfvbT33fa+3Peatrf1b2Dw2M4EBFX85x8Q5MfgsZ2dun5jbu/Lfa8P7YX3ZWdnUzklUQhRUYOeQXYkJDwQgJyTBViKLfj5+xHePLj0uLRvuO19ue+10b5kMVVJJYLwqOAaVSJozO19ue+eaK/8YOiUrrJATwjRoDTqABlMoBESHkheXh4HkrL4+q2tNfqlUb69rZ38wsPDycnJcbl9TfofGhrK7bffzkMPPUT37t1r1D43N5ewsIq5r/7+/vTq1YuioiI6derE3Llzad68ud3rFxQUMOvDD7nnnntc7n9NudO+tq5ta+zq8vqutu8ysA1dBrZxecFEbbV3dcGFN/svY2far/h0I9tXZ9O6Y0SNzyGEEPVZg0+xcNbujRmsnpdamltXUrpox7ojXu5ZRSVbhNrz3nvvOR0cOxIaGkpiYiLJyclERUXx5ptvVvv6kydP8tZbb9XKtYUQ9V9ke5PSk5YqJcOEEA1Lo5lB/mnBDo4dsD+Te2RvFpaiimWwis5YWDU3hd//d8hmm5ax4Vw0sUuN+5KRkcFdd93F/v37AZgxYwZ9+/Zl/fr1/OUvfymdEZw9ezZdu3Zlzpw5fP311+Tn53P69GmeeeYZnnvuOVq2bElycjL9+vVj5syZAAwdOpSXX36Z/v37Ex4ezgMPPMDSpUsJDQ1l8eLFtG7dmt27dzNlyhSKi4sZNWoU06dPJz09vdo+Dx48mKSkJABycnIYO3YsJ06coLCwkBdeeIGRI0fyxBNPsHv3bhISEhgxYgQvvfQSL730EgsWLKCgoIDx48fz/PPP13i8hBD1U2iEP02jQjiQeoKeF8d4uzvCSTExMRw/ftzb3RCiXvPoDLJS6nKl1Hal1C6l1BM2ng9WSn1qfX6dUqpjuefilVJrlVK/K6V+U0qFeLKvlYPjEsV2jrvjgQce4MEHH2TDhg188cUX3H777QDExcWxZs0atmzZwrRp03jyySdL26xdu5YPPviAVatWAaYO4IwZM9i2bRt79uxh7dq1Va5z+vRpBg0axNatWxkyZEhpEP3AAw/wwAMPsGHDBtq2beuwv8XFxaxcuZIxY8YAEBISwsKFC9m8eTOrV6/m4YcfRmvNP//5Tzp37kxiYiIvvfQS3333HTt37mT9+vUkJiayadMm1qxZ4/b4CSHqB6UUMXGRHNx+AovUlPcZl156KfHx8d7uhhD1msdmkJVS/sCbwAggDdiglFqitd5W7mW3ASe01ucopa4D/gVMUkoFAB8BN2qttyqlWgCF7vTH0UzvnL/+j9MnzlQ5Hh4VzPiH+7pz6SpWrFjBtm1lw5Cdnc2pU6coKCjg5ptvZufOnSilKCws+5ZHjBhBVFRU6dcDBgwgJsbM2CQkJPDHH39UuU5QUBBXXXUVAP369eP7778HTLC9aNEiAK6//noeeeQRm/3My8sjISGBffv20a9fP0aMGAGY2rpPPvkka9aswc/Pj4MHD3L06NEq1RK+++47vvvuu9IcyZycHHbu3MmQIUNqNmBCiHorplskKb8c5tiBU0R3aObt7gghRK3w5AzyAGCX1nqP1voMMB8YW+k1Y4EPrJ9/DgxXJsoaCSRprbcCaK2Pa62LPdhX+l/Rvs5KF1ksFtauXUtiYiKJiYkcPHiQpk2b8vTTTzNs2DCSk5P56quvKuwA2KRJxU1fgoPLynn5+/vbzE0ODAwsDVrtvaY6JTnIf/zxB2fOnCnNQZ43bx4ZGRls2rSJxMREWrduTUFB1bqoWmv++te/ln6fu3bt4rbbbqtRH4QQ9VtMV/OH+4EUyUP2FQsWLODnn3/2djeEqNc8GSC3Aw6U+zrNeszma7TWRUAW0ALoAmil1LdKqc1KqcdsXUApdadSaqNSamNxsXvxc+f+rRg2Ja602H14VDDDpsR5pHTRyJEjeeONN0q/TkxMBCArK4t27cwQzZkzp9avW2LQoEF88cUXAMyfP9/h6yMiInj99dd5+eWXKSwsJCsri+joaAIDA1m9enXp7HXTpk05depUabvLLruMWbNmlVbxOHjwoMNcZyGEbwlrFkSLduGkpZ7wdleEk3Jzczlzpuo7pkKIMp5cpGdrh47KSWr2XhMAXAicB+QCK5VSm7TWKyu8UOt3gXcBQkND3U6AKyldVJtyc3NLUyEAHnroIV5//XXuvfde4uPjKSoqYsiQIbz66qs89thj3HzzzUyfPp1LLrmkVvtR3owZM7jhhht45ZVXuPLKK4mIcFyiqU+fPvTu3Zv58+czZcoURo8eTf/+/UlISCAuLg6AFi1acMEFF9CzZ09GjRrFSy+9REpKCoMHDwZMybuPPvqI6Ohoj31vQoi6FxMXSfKPByk6U0xAkL+3uyOEEG7zZICcBsSW+zoGqFwOouQ1ada84wgg03r8R631MQCl1DKgL7ASH2OxVN0OGODTTz+t8HVeXh6DBw9mx44dpcf+9re/ATB16lSmTp1aenzo0KEMHTq09Os33nijdFekH374ofR4+frLEyZMYMKECQC0a9eOX3/9FaUU8+fPp3///jb7WLl+81dffVX6eeVFgSXX//jjjyscL1kQKIRouGLiItm68gCH92QRGxfluIEQQtRznkyx2ACcq5TqpJQKAq4DllR6zRLgZuvnE4BVWmsNfAvEK6XCrIHzxcA2RK3YtGkTCQkJxMfH89Zbb/HKK694u0tCCB/W9tzm+Pkp0lIkzUII0TB4bAZZa12klLoPE+z6A7O01r8rpaYBG7XWS4D3gblKqV2YmePrrG1PKKWmY4JsDSzTWn/tqb42NhdddBFbt26tcCw3N9dLvRFC+LqgkABan93MumFI7S9sFrWrU6dOZGRkeLsbQtRrHt0oRGu9DFhW6dgz5T7PB6610/YjTKk3IYQQ9VxMXBQbvt5L/ulCQpoEers7ohoXX3wxKSkp3u6GEPWabDUthBDCbbFxkaDh4A5JsxBC+D4JkIUQQrgtulMzAoP9JQ/ZB8ybN092NRXCAQmQhRBCuM3f3492XZpzIFU2DKnvCgsLcXfvACEaOgmQy0taAK/2hOeam49JC9w63dChQ/n2228rHJsxYwb33HNPte3Cw8MBOHToUGlpNlvn3rhxY7XnmTFjRoXFd1dccQUnT550putO6d27N5MnT66189WlqVOn0qlTJ9555x0ACgoKmDRpEueccw4DBw5k3759Ntu99tpr9OzZkx49ejBjxozS44mJiQwaNIiEhAT69+/P+vXra9SfNWvW0LdvXwICAvj888/tvm7Tpk306tWLc845h/vvvx9T9AUeffRR2rRpw8svv1yj6wpRm2LioshKz+NUZr7jFwshRD0mAbKV/+9fwFf3Q9YBQJuPX93vVpA8efLkKjvVzZ8/3+mgsm3bttUGS45UDpCXLVtG8+bNXT5feSkpKVgsFtasWcPp06dr5Zy21HR77Jp46aWXuOuuuwB4//33iYyMZNeuXTz44IM8/vjjVV6fnJzMzJkzWb9+PVu3bmXp0qXs3LkTgMcee4xnn32WxMREpk2bxmOP2dz80a727dszZ84crr/++mpfd/fdd/Puu++yc+dOdu7cyfLly6t8L0J4S0xcJIC1moUQQviuxhMgf/MEzL7S7iPwmwehMK9im8I8WHyf/XbfPFHtJSdMmMDSpUspKCgAYN++fRw6dIgLL7yQnJwchg8fTt++fenVq1eFTThK7Nu3j549ewJmI47rrruO+Ph4Jk2aVLoxB5ig6YILLqBHjx48++yzALz++uscOnSIYcOGMWzYMAA6duzIsWPHAJg+fTo9e/akZ8+epTOh+/bto1u3btxxxx306NGDkSNHVrhOeR9//DE33ngjI0eOZMmSsvLWu3bt4tJLL6V379707duX3bt3A/Dvf/+bXr160bt3b554woxb+VnwY8eO0bFjR8Bss33ttdcyevRoRo4cWWWsFi9eXHq9Dz/8kPj4eAYOHMiNN97IqVOn6NSpE4WFhQBkZ2fTsWPH0q/tWbx4MTfffHPpz23lypWls7MlUlJSGDRoEGFhYQQEBHDxxRezcOFCAJRSZGdnA2bL8LZt21Z7vco6duxIfHw8fn72/0kePnyY7OxsBg8ejFKKm266iUWLFtXoOkJ4UlTbJoQ2C+KA5CELIXycR8u8+ZRiO/vSFxe4fMoWLVowYMAAli9fztixY5k/fz6TJk1CKUVISAgLFy6kWbNmHDt2jIEDB3LttdeilK3dt+Htt98mLCyMpKQkkpKS6Nu3b+lzL774IqGhoQQFBTF8+HCSkpK4//77mT59OqtXr6Zly5YVzrVp0yZmz57NunXr0FozcOBABgwYQNu2bdm5cyeffPIJM2fOZOLEiXzxxRfccMMNVfrz6aef8v3337N9+3beeOMNxo0bB8CUKVN44oknGD9+PPn5+VgsFr755hsWLVrEunXrCAsLIzPT8ezS2rVrSUpKIioqiqKiogpjNWjQIMaMGcO2bdt48cUX+fnnn2nSpAl5eXk0bdqUoUOH8vXXXzNu3Djmz5/PNddcQ2Bg9WWnDh48SGys2fgxICCAiIgIjh8/XmHsevbsyVNPPcXx48cJDQ1l2bJlpbsQzpgxg8suu4xHHnkEi8XCL7/84vB7rKmDBw9W2LY8JiaGgwcP1vp1hHCVUorYuEgOpJ5Aa233/zPhXV26dOHo0aPe7oYQ9ZpTAbJS6kLgXK31bKVUKyBca73Xs12rZaP+We3TenoPVHZa1SciYuEW1/coKUmzKAmQZ82aZa6nNU8++SRr1qzBz8+PQ4cOcfToUdq0aWPzPGvWrOH+++8HID4+nvj4+NLnFixYwDvvvIPFYuHw4cNs27atwvOV/e9//2P8+PE0adIEgKuvvppffvmFCRMm0KlTJxISEgDo16+fzVzcDRs20KpVKzp06EBMTAy33norJ06coKioiIMHDzJ+/HgAQkJCAFixYgW33HILYWFhAERFOd6KdsSIEaWvqzxWBw8e5OjRo6xatYoJEybQsmVL8vLySl9/++238+9//5tx48Yxe/ZsZs6c6fB6lWeLgSq/3Lt168bjjz/OiBEjCA8Pp3fv3gQEmH9Cb7/9Nq+++irXXHMNCxYs4LbbbmPFihUOr1sTzvRRCG+LiYtkx/qjZB46TYt24d7ujqgsaQHnr5uGzkqDfU/D8GcgfqK3eyWclbQAVk4jLisNImLk5+dBDlMslFLPAo8Df7UeCqQBbuBRNOSvEBha8WBgqLn53DBu3DhWrlzJ5s2bycvLK535nTdvHhkZGWzatInExESio6PJz69+YYutYGjv3r28/PLLLFu2jKSkJK688kqH57EVaJUIDg4u/dzf399mDvAnn3xCamoqHTt2pHPnzmRnZ7No0SK757U3kxQQEIDFYgGo0ueS4B2qjlXr1q3Jz8+3e94LLriAffv28eOPP1JcXFyaplKdmJgYDhw4AJi856ysLJuB/G233cbmzZtZs2YNUVFRnHvuuQB88MEHXH311QBce+21Nhfp3XLLLSQkJHDFFVc47I+9Pqallf0Rl5aWVuNUDiE8LSbO/LtJS5U0i3onaUHpWhtVS2ttRB2Sn1+dciYHeTwwBjgNoLU+BDT1ZKe8objHNTD6dTNjjDIfR7/u9l9m4eHhDB06lFtvvbXC4rysrCyio6MJDAxk9erV7N+/v9rzDBkyhHnz5gFmsVhSUhJgcmybNGlCREQER48e5Ztvvilt07RpU06dOmXzXIsWLSI3N5fTp0+zcOFCzj//fKe+H4vFwmeffUZSUhL79u1j3759LF68mM8++4xmzZoRExNTmhdbUFBAbm4uI0eOZNasWaULBktSLDp27MimTZsAqs2lrTxWf/zxBwDDhw9nwYIFHD9+vMJ5AW666SYmT57MLbfc4tT3NWbMGD744AMAPv/8cy655BKbwXd6ejoA+/fv58svvyz9mbZt25Yff/wRgFWrVpUGzuXNnj2bxMREli1bVuU5Z5x11lk0bdqUX3/9Fa01H374IWPHjnXpXEJ4StOoEJq3DpNyb/XRymm219qsnOad/jRG1mpZCYuHOV8tq6gATh6A756y/fNb8VyNrx/36eBaqdZVI968tgucSbE4o7XWSikNoJRq4qiBz4qf6JG3KiZPnszVV19doaLFlClTGD16NP379ychIYGuXbtWe467776bW265hfj4eBISEhgwYABgSq316dOHfv360blzZy644ILSNnfeeSejRo3irLPOYvXq1aXH+/bty9SpU0vPcfvtt5OQkFAa/FVnzZo1tGvXjnbt2pUeGzJkCCkpKRw+fJi5c+fypz/9iWeeeYbAwEA+++wzLr/8chITE+nfvz9BQUFcccUV/P3vf+eRRx5h4sSJzJ07l4suusjuNSuPVVxcHAA9evTgqaee4uKLL8bPz4++ffsyZ86c0jb/93//53TFkNtuu40bb7yRc845h6ioqNKf1aFDh7j99ttLg9prrrmG48ePExgYyJtvvklkZCS5ubnMnDmTBx54gKKiIkJCQnj33Xedum6JDRs2MH78eE6cOMFXX33Fs88+y++//w5AQkICiYmJgEnlmDp1Knl5eYwaNYpRo0Y5fMdAiLoWExfJ9l+PUFxswd+/8awFrxPOvsVusUDmHjj6Gxz5DY4kW6s02ZBlI71Q1L6SGeDCPBSYn8eS+8zPplVXyDkKOelwOt18zDlqHvlZ1Z83+yC82BbCo8s9Wlsf0dDE+nXaelj5fMXrf2VSN52KfdxJ77D1vdfk2l6gqnu7HUAp9QhwLjAC+AdwK/CJ1vp1z3fPeaGhobpyxYWUlBS6devmVPu8vDxCQ0Mdv7CBts/NzS3NEfbG9Wu7/eeff87ixYuZO3euzddPnTqVq666igkTJjSYsXvuuecIDw/nkUceqfKa6v4t1OTfiSfab9myhT59+njt+u60l7Gz3X73lnSW/zeZqx/py1nn2C8tKeNXw/blgoxSgaEw6iVoFQdHkuBosgmIj26DQmsJTuVvArAT+6Awt+p5m7aFh1Nq1HefG7v60H56dxPMVieoqY1A1xrkrpwGuceqtgmJgIQbrIH10bLgOs/JNKegcLjo4XJBdSvzMawl+FvnUe3de+XfaS86U6kP5QL9LR9BkY2qWKEt4JZl0OKcsms54O69p5TK1Vo7nOx12But9ctKqRFANtAVeEZr/b3LPRPCw/785z/zzTffVJvKEBERwdNPP82xY8dKy7v5skcffZSFCxfy8MMPe7srQtCuSyQoOJB6otoAWdSQvRSJJfeVfR0cAW16Qt8boXVPaNPLBM+BIbaDHIAzuXAoEdomeP57aIyLzHIyYN071QTHCu7fYgLhoGritsBQ20HqFS/bHsOiM3A6oyxg/WSS7fOeyTEzy7b61aSlCZaP7zSpHuWVlMJd87J1ptvORmQhzW0HxwB5x+GtgRAQAtHdzP3aJt7cu617QEizstda752ErDT4wfP3jsMAWSn1L63148D3No4JUe/85z//cfia1157rfRze7WefclLL73ESy+95O1uCAFASJNAots3JS01kwFXdfJ2d+ofZ4LE4kLI2F42I3zkN/spEgDXfWyCiubtwV51G+s1chY/RpPiE6iIWOh/C2yYBbMuh6v/C909uK7BB99md0vmXvjlP5A4zwSXAaG2A8WIGIhy4t9JyRitNFVIlKM/MAKCIKKdeYBZW2XrHoqIhXvXWdM7MspSO0rTPDLMfWhLcYF5d6LTENuz3uHREBBsco5tXTu8NVz6vPU+T4KUpbD5w7LnIzuaoBkFO5ZD8Zk6u3ecmc8egaliUd4oG8fqJanFKRo7R2lUQnhCTLcoEr/bz5n8IoJCpOR+KZt5qPdDRqp5S7skUMjYXlaf3z8YWneHwCZlaRPlRcRC3JXOXT9+IikFnThy5AijR482xxJugE+nwIKb4JL/g4sesR9ku8PeDPg3j0N0dxNo+Vdfs94nHEqEn1+DbYvALwB6Xwfn3w+HttieAa5JtSzrWqlUV1I8hj9j//pBTUyQbi9QtxfgRsTCJNupjE5de+QLFQNcrSH7UMU/DI8mw/FdVc9ZssC0rgNkpdTdwD3A2UqppHJPNQV+9khvallISAjHjx+nRYsWEiSLRklrzfHjx0trUgtRV2LiItm8/A8O7TxJx14tHTdoDCwW+P6ZqkFiUR789Ir5vEm0SZHofAm07mVmz0ryM+3lgdawHOl5551HSkq5nOOmreHmpbDkz7DqBROcj/lP1dKn7sg+ZH8GPC8T3rkA/INMkFzyFnubXmYsQiPLXltfUzS0hj0/wM8zzMfgZnD+n2Hg3dDsLPOaltbqRs7OANe2ms5Al1ddcF2b11aqbNa7y2Vlx59rDtiY7PHgAtPq/qz/GPgGszCv/J7Kp7TWPlG/p6RubEZGhsPXFhYWOtxtrSG3P3PmDEFBQV67vjfbN/SxCwkJqbADnxB14azOEfgH+pGWeqL2A2R3gyR3cxmduf6ZXEhPqVhF4miyyfe0ScHD202wao87AU45hYWFVWvcB4bA1e+aAHXV30x6wHUfV98fZxTkmNnUX6pJfQtvA5e9aGbOjyTDzu9NWkKJiFgTMCs/2PW9e2+z19K9U9p+2P9BQKD5Hg9vNd/Lpc+b1JWQiKrtrTPAiW4uNHOZqzPQtXHvuTP7HRFjZwbbc7/b7AbIWussIAuYDKCUigZCgHClVLjWuvrCvfVAYGAgnTo5l//mkytia7H9li1bfLr/3qxE4OtjJ4QnBAT6c1bnCNJqux6yu3msnmi/5M8msAttXvbW8PFdoM1GSAQ1NTOhCdfDb5/Zri4QEeNcMOpOkGE1b948cnNz6dWrV8UnlIIhj0DLLrDwTzDzEpj8CZxlf2dWuyzFsGUurHrRVDboeQ2062eC7ypvs/8Nek0wjxKnjlr/uEguG9OM1KrXKcwz47/nx3L5r60qljkLbma+N0/87BfdBWgzyz/6dZNOERDs4EQ+qhbuPZe5O4PtAmcW6Y0GpgNtgXSgA5AC9PBYr4QQQvi82G5RrF24m9zsM4Q1c/1dllJa205RKMyDpQ+ZGTxHNn1Q++2L8uEX68LfiPYmNaDHePOxdU9o3gH8rPWgY86r81/0NdZ9DER2gE8mw6zL4OqZ0O0q59vvXAHfPw3p2yB2kAmyY/qb55q0cm4Wsmlr8zjn0rJj9t5mL8qH3atMIG6puvsr/sEmWM45UpbXXcLdnz0awlrAvRvKfsai9tXSuyc14czKiReAQcAKrXUfpdQwrLPKQgghhD0xcSZ3NG17Jl3Oa1P2hDNvcxfmWVMUkiumKBRk277YmVOwaY7jTtlLcXC3PQoe31sxX9YWL/yid8lZveGOVTB/ilnAN/wZuPCh6hfvHUk2gfHuVRDZCSZ+CN3GVGzjkbfZY+HBZJPjnXfCVF6osNmG9fOk+VXbgvs/+9xMCY7rQh2npzgTIBdqrY8rpfyUUn5a69VKqX95vGdCCCF8WsvYpgSHBZCWcqIsQLaXonD0dxNclgTEx3aCLjZtgsJNTdT4iZD8hZ0UBWuQ5Eh1q/Hdah/jODgu4c23qmuiaRuYutTUul05zSzeG/26yVcuL/swrH4BtswzebeX/QPOu92UGatNjt5m9/ODJi3Mg+5V2//xs+d+9qLBcSZAPqmUCgfWAPOUUumAjfcwhBBCiDJ+foqYrpEcSMksK7lpq9RXUb5Z/Q/QLMakJnQbXZaiENmpbIYudqB7KQru5jJ6IRfSqwJD4Zr3zGYjq1+AAxugOJ+E7MOwui207Qu7V5q6zYPvNTnMzv6hUFPuzr7Lz17UgDMB8lggD3gQmAJEANM82SkhhBANQ0y3KHZvySArPY/mrcOqKcuk4LE9EBZV/QndDZK83b4eSEhI4NChQ843UAouftRsc7zuHXMIzM5w2QehbT9KKe2QAAAgAElEQVSY8L5zm124y53Zd/nZixpwZqvpkqrkFuADpZQ/cB0wz34rIYQQAmK6luQhnzABckQ720FyRIzj4LiEuykK7uYy+kqKhB0JCQkEB7tQaSH1a9vHT6fXTXBcG2rp3vHVn71wnt2scqVUM6XUX5VSbyilRirjPmAPIH8uCSGEcCgiOpTwqGDSUqzl3jpcWPVF8jZ1ncrNzaWgoKDmDe3N/ntwswYhvKW6ZZdzga7Ab8DtwHfAtcBYrbUHN2sXQog6kLQAXu1J3KeDzeKbpAXe7lGDpJQiNi6KtO0nsJw4AKlLzbbCEbFolFkgNfp1eZu6Di1YsIBffvml5g3tLUaTRWqiAaouxeJsrXUvAKXUe8AxoL3W+lSd9EwIITzF3Q0DRI3EdIsk5ZdDFH75F4K1xdTFjewob1P7GlmkJhqR6maQC0s+0VoXA3slOBZCNAi2KikU5sHK573TnwYupmsUnYPXEnxgBQx7EiI7ertLwhXxE81sv8z+i0aguhnk3kqpkorsCgi1fq0ArbVu5ujkSqnLgdcAf+A9rfU/Kz0fDHwI9AOOA5O01vvKPd8e2AY8p7V+2envSgghqlNdLuV/h1jLi/WyfuxhthAuz5mNLkSpsMBcLo58n5P+59J84N3e7o5wRx1v1iCEt9gNkLXW/u6c2Frt4k1gBJAGbFBKLdFabyv3stuAE1rrc5RS1wH/AiaVe/5V4Bt3+iGE8BBfDRK1hsAwKDxd9bngphAaBduXw5aPyo43b18WMBdkwcbZUJTvenqGr46dq1Y8Swgn+frYU4wrVgS49dtFCCE8z5k6yK4aAOzSWu8BUErNx9RULh8gjwWes37+OfCGUkpprbVSahymYoaN32JCCK/y5Rzede+Y4NgvACzl9jwKDIUrp5v+a222qD3yW9njaDLs+Aa0peo5C/Ng2SNQVADhrSE82nxs0hL8Ayu+1pfHzhV//AKb5pB97m0c/elsDu/JIjbOyXJuwiP69+/PwYMHvd0NIeo1TwbI7YDyezKmAQPtvUZrXaSUygJaKKXygMcxs8+P2LuAUupO4E6AwMBAey8TQtQ2uzm80+p3kLdzBXz7JMRdBd3GwKq/2S74r5TZZrdpGzh3RFn7M7nw97aArnru/CxYcl/V42EtrMFyK/Nx+ze+OXauKCqArx6AiPaEjn4Wv583k5Z6QgJkL+vZsyf+/jKNL0R1PBkgKxvHKv9Wsfea54FXtdY5Stl6ifWFWr8LvAsQGhpq4zeWEMIjfLEeasYO+PwWiO4B4/8LweHQe1LNKikEhZmUiKwDVZ9rFgO3LIOcdDP7fDq97PMc6+cH1sEZO2ud63Ls6irF46fpcGwHTPmCoGYRtD67mamHPK5z7V9LOC0rK4vc3Fxvd0OIes2TAXIaEFvu6xig8t6WJa9JU0oFYLaxzsTMNE9QSv0baA5YlFL5Wus3PNhfIYSzQiMhL7Pq8fpaDzU3Ez6ZBAHBMPljExy7yl6pq0ufhcgO5lGdV3vaDrADQ+HkAWgeW/W52lRXKR4Z2+GnV6DXtXDupQDExEWx4eu95J8uJKSJvOvnLQsXLiQ3N5d+/fp5uytC1FvVlXkDQCl1SimVXelxQCm1UCl1djVNNwDnKqU6KaWCMNtTL6n0miXAzdbPJwCrtHGR1rqj1rojMAP4uwTHot6xbjSRsHhY49poYv1MExwrG/99tO1n8nfrk+JC+GyqmaGdNM8suHOHu6Wuhj9jguHy/AKg6Az8px+seA7ys202rRXVpcfUFovFpFYENYHL/lF6ODYuEjQc3HGi9q4lhBAe4DBABqYDj2LyhWMwOcEzgfnALHuNtNZFwH3At0AKsEBr/btSappSaoz1Ze9jco53AQ8BT7j6jQhRp0pm4bIOoNBls3B1GSTX9U5wWsOP/zaL0bpeAWP+UxYkNouB2EGQsgi+vKNqAOZNy5+AvT/CVTOgfeVlEC6KnwgPJpM6aS08mFyzmVdbAfa4t+GBROgxHv73Krzex/whUlzo+Hw1kZ9le/YaajfFY/Mc2L8WLnsRwluVHo7u1IzAYH/SUiVAFkLUb86kWFyutS7/W+VdpdSvWutpSqknq2uotV4GLKt07Jlyn+djtq+u7hzPOdFHIeqWtxep1XUlBIsFvnsKfn0Lek+GMW+AfwD0uaEsh1dr+N90MwaZe+G6j6Fp69rvS02snwkb3oPz/wx9pni3L+VZa8lWyX+++r8w6C749v/MHyLr34UR06DL5WbhoCssxbBnNSR+YrZ5tqdpG9fOX1n2Yfj+Weh4ESRUHHN/fz/admkuAbIQot5zZgbZopSaqJTysz7K//atZ++lihprrGkC7vL2IrW6eJu8RHERLL7XBMcD74axb5nguDKl4KKHYdJHkL4NZg6Dw1trvz/O2vMjfPM4nHsZXOpDO+S17QNTl8J1n5iScp9cBx+MhkOJNTtPegp89zRM7w4fXQO7VkCfG2Dok1VTPMCkdez41v3+L3/cVK8Y/ZrNoD42LoqTR3M5lZnv/rWEEMJDnAmQpwA3AunAUevnNyilQjEpFMJXNcY0gdqgtf1FXs3Oqps+1FWAXpgPC26CrR/DsKfg8n+An4P/NrqNhluXm89nXQ4pX9Vun5xxfLfpd8tz4Zr3wM/HSlopBXFXwD2/whUvmz843r0YvvyT+Rnb+8P29HFY91/478Xw1iBY+6YJuCd+CI/sgCtfgaGPV03xGPECRJ0NH0+Eb58y+dCuSF0G2xbDxY9BC9uVKmLiIgFIS7WxyFPUicGDB9O1a1dvd0OIes1hioV1o4/Rdp7+X+12R9SpxpYmUBssxbD0QSg4ZYIuS3HF57UyQUqTFp7tR5MWcPpY1eNBYZCTUSHv02X52TD/etj3kwnSBtzhfNuzesMdq037T28wC9MufMj1NIGayM8ys67KDybPh5Bmnr+mp/gHmnGPn2hKpv36Nvz2uSmQaSkq+3ez+F4TDB/9HSyFZse/y/5hKkjYuhdspXgMuMOk0ax9w+QPT5gFkR2d72t+tkkLie4O599v92VRbZsQ2iyItNQTxAx2Zo5G1LauXbtisdjY8EYIUcqZKhatlFJPKqXeVUrNKnnUReeEhzWmNIHaUFQAn98Kmz+AIY/CuHcqzsJd8CDkHYe5Y01ZMU9J/hJOZ1KljLhfgNnI4vU+pryWOwvlTh8zb+vvXwtXv1ez4LhE09YmVaDnBPMzXfgnMyPtScVF5meUucfMmkZ18uz16kpIBIx4Hu7bAAFBFXcABCg+A0eSYOCf4K6f4a7/weB7avaHUmCImWG+9gM4thPeGWJmg5216gXIPmRmpwOC7L5MKUVM10gOpJ5A17eKJ43EsWPHyM72YKUUIRoAZxbpLQZ+AlYAxQ5eK3xF4sfYTSGvq1q23g7Qa6Igx8yE7lkNl/0dBt9rjsdPJHHLFvr06WO+7nQhfDIZProablwEoc1rtx8bZ5sZ7NiBZrHcTy9X3AmubR/4/hkTkG6YZY71utZxWkR5WWkwdzyc3G8W2nW5zPX+BoaaFIfoOBNAZe4x5wyPdv2c1fn+GZNrO/o16HSRZ67hTZEd7P/ho7WpGuGuHuPMOwCf32rSVM67A0a+YAJoe9I2mgWFA+6A2PMcXiIw2I+87DOsfv0IG6JOMHhsZ7oMdH6R4I51R1i7eDc5mQWsj8p0q31S1M9evX5N29fGtVfMS0Kf8Sd1sXvfuzfGToi64kyAHKa1ftzjPRF1w1IMK5+Hn1+Dll1NEFRU6Rdu7ADP96Mgx8wyFRVUfa6+bTaRm2lyMw9uMgvUqquGcM6lMHGuCaY/ugZuXFg7b/Frbcp/rXwezhlhZkeDwqD/1KqVECZ/Ant/Mm+XL7zTLK677EXoeKHj6xzbCR+Og4JsuOFL6HiB+31Xysy4t+xicmjfHQbXzzdpALVp84fw65sw8C7oN7V2z12f2NvJrzb/3UR1glu/Nffb2jfMDoDXzrGdV1xcCEvuh6ZnwSVPOzz1jnVH2L7uaOnXOZkFrPooldNZBXTq7XjGe+/WDNZ9tZfiQkuja19b19aFAXXed3vtV89LBZAgWdQ7zgTIS5VSV1hLtglfVpBjatRuXwb9b4VR/4bfF8LKadZZyHYQ3gaSv4B2/c1btJ6QmwnzrjXBsX+QeXu4hPKDi+tROezsw2Y2+PguE/h2u8pxm66Xm2Dis5vN93nDF+7t3KY1fP80/PIfk64w7u1q38IGzOzpHT/Ab5+Z2eQ5V5raxSOmmYVrthxKNEE9mNSIs3q73mdbuo+F5h3MDPv7l0G/myHlKxKy0uAHN7c7/uMXWPoQnD0MRtbCLGp9Zm8nv+HP2G/jioCgsj+sFt0N/x1iaknHV6rM+cvrkP67qbrhxB+DaxfvLg2QShQXWvjly9388uVul7ramNv7ct8Bis5YWLt4twTIot5xJkB+AHhSKVUAFGISH7XW2odXvjRCJw+YhUvp22DUS+atUKVKF+uUpgkUF5q3Vr/9a9kCodqUfdi8fZ+52+xqVphbFqCXbF/82wLodY3tUlR1KXOPmU3NPQ5TPoezL3a+bber4Jr3zVh+PBGmfGZ2FaspS7HZkWzLXDjvdvOzczZdws8Pek+C7mPMLPJPr8KbA80fR0OfgN2rYOU04rLSoElLs8gqPNqkhrQ8p+Z9dUbbBLhjFcy6zPQJXF+gmbSgrP9KQVhLuHa27RJ0DUnJ+JT+YevmHxeOdB1lcpo/vw2+vN1suhI7CH78pxl7NLTta6puOCEn08a7RlaX3tLdYfsVs7c12va+3Pfq2ld3TwjhLc5UsWhaFx0RHnRgvakmUHTGBGrnXGr/tf6BJrD77GazIt0vAPrfUjv9OL4b5o4zM8jlA87yq+kTPzGzVfOnmFzV6vIePelIspk5Li6Em5dAu341P0ePcWYx1Zd3mD9Orl9Qs6C/qAC+uB1SlpgUhWFPuVYFIjDU1CfucxP88A/YOAs2zwVdDJZCE6CezgAUDP6z54LjEs3Oqlr9A8yM6Ff3m3zW8GgIb219RJtHk1bm/oSqFVC0NmkhO7+vvxVQalPlP2w9LSIGpn4NP/zdLADd8hGY5alG+jbzM3Fi7MOjgm0GROFRwXR1YhbxV2v+amNs78t9d9ReiPrG7lSUUirO+rGvrUfddVG4Zeun5u31oHC4fUX1wXGJgCCTInDuSFj6FxNMuetIsqmJW5BjAk57s7EJk80WxrtXmgVCrtZjdcf+dTDnClD+pp6vK8FxiV4TTLWLvT+ZP1KcreJQkGNmnlOWmHJdl/yf+yXSwlvBVdPhnrXWMmGVtzHWsPY/7l3DWdkHbR8vzIOtn8Cqv8GS++Dja0393+nd4G8t4d9nw1uDYfF9VRerFeXX3wooDYF/gJmpbtKKKgt8azD2g8d2JiCo4q+egCA/Bo+1XTdZ2tePa3uqvZ+fcrq9EHWpuhnkh4A7gVdsPKeBSzzSI1E7LBZY/YKZ7el4kVnUFRblfPuAYJNzO38yLPmzmUlOmOxaX/avM4FOYBO4+StT0aA6fW80wdvSB+HzW0ywXjJz6Gk7V5gFds3awk2LoHl798/Ze5L5fhbfCwtuNDvNBVQzY1KSo31oi8k3Trje/T6U16qr7cWRUHcVROwuNIuFB5NN8JuTbn0chdPlPs9JNzOWttTHCigNja362+D02JfkmpZUMgiPCq5RJYPG3N6X+26rfUCQH8XFmphuNfjdJEQdsRsga63vtH4cVnfdEbXizGn48k5IXQp9bzabPDha1GVLYIhJc/h4Eiy+xwTJlRfoOOJqwNn/VlPT9ptH4Yvb4JpZns8tTf7CVFmIjoMbFtbOZhsl+txg0jWW/gU+m2pqzdr6mWQfsuZoW+v4OrMo0BV1UQmhOo4WmgWGmpJmkR1st3+1p3f735jVwr3TZWAbugxsQ0rlCix13H6Liykq3ux/bVy7WQfYs2cP559/vkvta2vszorqwMfPr2PjN/sYMqlLjc8lhCc5tdpHKXW+Uup6pdRNJQ9Pd0y4KCvNLIDavgwu/6epB+tKcFwiMNTsRtb+fLPRw+8LnW+b/IXJvW15jklVqOls7MA7Tc3hbYtNubLiIsdtaqL8Ntf/6mgW1MWcZ3ItazM4LtH/FvPHyvZl8MWtJmAu7/hu87PLSjM52p4KjsEEopXzoT1RCcGe+IlVtzse/brz+cPe7n9jJmPv85YvX05iYqK3u0Hz1mF0O/8sfl9zkOxjbmxsJIQHOJySU0rNBToDiZRtFKKBDz3YL+Gs8iv5w1vBmTyTq3r9Ajh3RO1cIygMrv/UlAD74nbwC3QcvG2cZcputR9sat6GRLh27cH3mkByxbNmBnvc22aLZ3dVXuSVd8KUmOszxfW+OmPAHWbh3vInTE52zhHisg6aRWhnck0qiauLAmuirish2OuDqwvN6kP/GysZe1GLzruyI9vXHWH90r1cOtVxJQwh6ooz71n3B7pr2RO0/qkc5OWkAwoufa72guMSweGmAsZHV5sUgUkfmXq/lWkN/5tuFuyce5nJHw4Kc+/aF/7F5PCuesEEyWPeqNnOcLZ8/2zVRV7aAj/806RDeNKgu02lhuTPAWuZs5yj5rMRz3s+OC5hDVCrbDTiK3y9/75Mxl7UkvDIEHoNjSFxxX76jGhPi3Zu1IwXohY5E2UkA1LBuz5aOc3GtrMaNrznmeuFNDObXrTpaRab7VxR6dIavvs/069e18J189wPjksMedRsIJI4z+TxWiyO21RWmA/JX5oFcKcO2X5NXS3yOrDOxkEN62fWzfWFEKKe6HdZB4KC/Vm3ZI+3uyJEKWdmkFsC25RS64HSpe9a6zEe65Vwjr1gzpNBXkiE2T75gzGmbNmguyH5C5PiERgGhafhvDvMLn3uzvJWNvQJM5P80ysmFeGKlx2XPtMa0jZA4sfw+5eQnwXN2kFwUyg4VfX1dbXIyxs/OyGEqIdCwgPpM7I965bs5cieLNqc7cE0NyGc5EyA/JynOyFc5K1KBKGRcNNieOci+HkGYE0TKDxtUiBiB9R+cAwmGL7kaZOT/MvrJhf68n/YDpJPHoCk+WbjkczdEBBqdpTrPRk6DTELCOtiu157vF1FQgjRaA0fPpx9+/Z5uxsVxF8SS9LqNH5dtJuxD/ZBuVv3XQg3VRsgK6X8gae11k7sLiHq3PBnzIYJxeVq2tZVkBcWBdhIc7AUmRQLTy3YUQpGTDNB8rq3TfCbnmJmsJu1NZubZO42G3OgocOFcNFD0H2smTUu4e2FRo7KnAkhhIfExsaSk5Pj7W5UEBQSQP8rOvLTpzs5kJJJ++4tvN0l0chVGyBrrYuVUrlKqQitdVZddUo4KX4i7FoFSZ+gUXUf5GUftn3c02kCSpmZ4/RU2PmdOQRmd7ZNsyGspUnH6H0dRHa0fx5vLjTydoAuhGi0Dhw4wLFjdjZ88aIeF7YjccUBfl20h9i4KJSfzCIL73EmxSIf+E0p9T1wuuSg1vp+j/VKOC8oFEIiSB2znG7d67hEjjfTBJSCzJ22nwsMNQFyfSeVAIQQXrBy5Upyc3O56KKLvN2VCvwD/RgwuhMr56Swe0sG5/SL9naXRCPmTKLo18DTwBpgU7mHqA/SU6FVN8eL1TzB2xsGZB20c1wWugkhhC/qMqANUW2bsG7JHizFLlQrEqKWOJxB1lp/UBcdES7QGjJSTH6tN3g7TUAWugkhRIPi56cYOOZsvnnnN1LXHqH7hW293SXRSDmcQVZKnauU+lwptU0ptafkURedEw7kpJsd4Fp58e35+InwYDKpk9bCg8l1m0Pr7RlsIYQQta5T75a07tSM9Uv3UnSm2HEDITzAmRSL2cDbQBEwDLPF9FxPdko4KSPFfIyO824/vCV+Iox+HSJi0SiIiDVfy0I3IYTwWUopBo/rzOmTBfz2o51UOiE8zJkAOVRrvRJQWus/tNbPAZd4tlvCKemp5qM3Z5C9zZsz2EII4YMuv/xyEhISvN2NarXrGkls9yg2Ld9HQV6Rt7sjGiFnAuR8pZQfsFMpdZ9SajwgS0vrg/RtZtOOcPlxCCGEcE6bNm2IjIz0djccGjT2bApOF5H4/X5vd0U0Qs4EyH8BwoD7gX7ADcDNzpxcKXW5Umq7UmqXUqpK3S2lVLBS6lPr8+uUUh2tx0copTYppX6zfpQZa1syvFjBQgghhE/as2cPR48e9XY3HIru0IzOfaNJXHmA3Owz3u6OaGQcBsha6w1a6xzghNb6Fq31NVrrXx21s+7C9yYwCugOTFZKVS7Ue5v1vOcArwL/sh4/BozWWvfCBOOS81yZ1ibForHmHwshhHDJmjVr2LZtm7e74ZSBYzpRXGhh0zf7vN0V0cg4U8VisFJqG5Bi/bq3UuotJ849ANiltd6jtT4DzAcq1yMbC5SUkfscGK6UUlrrLVrrQ9bjvwMhSqlgJ67ZeJw6DAVZEF3Hm4MIIYQQdSSyTRO6DW5D8pqDZB/L83Z3RCPiTIrFDOAy4DiA1norMMSJdu2A8kVq06zHbL5Ga10EZAGVN2C/BtiitS6ofAGl1J1KqY1KqY3FxY2sFEy6tYJFK5lBFkII0XCdd1UnlFJsWLrX210RjYgzATJa68q7MTgTjdpKjNU1eY1Sqgcm7eJPdvr1rta6v9a6v7+/vxNdakAyrBUsohtxBQshhBANXnhkCL2GtiN13RGOH8rxdndEI+FMgHxAKXU+oJVSQUqpR7CmWziQBsSW+zoGOGTvNUqpACACyLR+HQMsBG7SWu924nqNS3oKhLWEJi293RMhhBDCo/pd3pGgYH/WLZZ9ykTdcCZAvgu4F5MOkQYkAPc40W4DcK5SqpNSKgi4DlhS6TVLKKuIMQFYpbXWSqnmwNfAX7XWPztxrcYnI1Vmj4UQQtTYVVddRb9+/bzdjRoJCQ8kYUR79m49xpE9Wd7ujmgEAhy9QGt9DJhS/phS6i+Y3OTq2hUppe4DvgX8gVla69+VUtOAjVrrJcD7wFyl1C7MzPF11ub3AecATyulnrYeG6m1Tnf+W2vAtIaM7RA/yds9EUII4WNatmxJRkaGt7tRY72Hx7Lluz9YNH0zxUWa9VGZDB7bmS4D2zh9jh3rjrB28W5yMgu80l74DocBsh0P4SBABtBaLwOWVTr2TLnP84FrbbR7AXjBxb41fNkHoSBbSrwJIYSose3bt3Po0CG6dfOtdyH3bT1GcZHGUmyWKuVkFrDqo1TyTxdydp9WDtvv2ZLBLwt3U1xoqbX2q+eZ9UASJDc8rgbIsjOFN5VWsPCt/9yEEEJ439q1a8nNzWX48OHe7kqNrF28uzQ4LlFcaOGnBTv5acFOl87pbvuiMxbWLt4tAXID5GqAXLkahahLJQGy5CALIYRoJHIyq1R7LTXsBsfvqK7+KNUj7avrl/BddgNkpdQpbAfCCgj1WI+EYxmpEN4awqK83RMhhBCiToRHBdsMRsOjgul+YVuH7Tcs2+uR9k2ayz5mDZHdKhZa66Za62Y2Hk211q7OPIvakJ4iG4QIIYRoVAaP7UxAUMWwJSDIj8FjO3utPYC2WMg5IbPIDY1TG4WIesRiMRUsJL1CCCFEI9JlYBuGTYkjPMrM2IZHBTNsSpzT+b+eaN//ig4UnrGw8JVNZGXIVtgNicwE+5qsA1B4WmaQhRBCuGT8+PHs2rXL291wSZeBbegysA0pKSkuVeHwRPtOvVux5PVEFr68iTEP9CGqbZMan1fUPzKD7Gtki2khhBBuiIiIICwszNvdaDCiOzRj/EN90RoWTt9Mxv5T3u6SqAUSIPua0hJvMoMshBCi5pKTk9m/f7+3u9GgtGgXzvhH+hIY5M+i6Zs5tOukt7sk3CQBsq/JSIWmZ0Foc2/3RAghhA/auHEju3fv9nY3Gpzm0WFc/WhfwiKC+eq1RPZvO+7tLgk3SIDsa6SChRBCCFEvhUeGMP7hvjRvE8bXbyWxZ4vvbektDAmQfYlUsBBCCCHqtbBmQYx7sA/R7ZuyfGYy23897O0uCRdIgOxLTu6DojwJkIUQQoh6LDgskNH3J9CuS3NWzEnhtx/SvN0lUUMSIPuSdGsFi1YSIAshhBD1WVBIAFfeG0/H+Jasmb+DTcv3ebtLogakDrIvySipYNHVu/0QQgjhsyZOnMiOHTu83Y1GISDQn8v/1JOVc1L4ddEeDu/J4nhaDjmZBSRF/czgsZ2d3qgEYMe6I6xdvJuczALWR2XWefvGRAJkX5KeCs1iIKSZt3sihBDCR4WFhREcHOztbjQa/v5+XHpLd05nF/BHUllli5zMAlZ9lEpBXhGd+0Y7PM/uzen8/MUuigsttdZ+9TzzzrQEyVVJgOxLMlIgWipYCCGEcF1iYiKHDh1yaSc54Ro/P0W2ja2oiwstrJm/gzXzXZvRd7d90RkLaxfvlgDZBgmQfYWlGI7thE4Xe7snQgghfFhiYiK5ubne7kajk5NZYPe5iyd3cdj+x0/sB8HutM/JLEBrjVLK4TkaEwmQfcWJfVCULxUshBBCCB8UHhVsM0gOjwqm58UxDttv+vYPj7QH+PTFDfS7rAOd+7bCz1/qN4BUsfAdpVtMS4AshBBC+JrBYzsTEFQx7AoI8mPw2M7eax/oR/cL22IpsvDd+78z77l1JK85SFFhsVPnbMhkBtlXSAULIYQQwmeV5PmWVJEIjwquURUJT7bXFs2erRlsXv4HP368nQ1f76X38Fh6DmlHUEjjDBUb53fti9JToHl7CA73dk+EEEII4YIuA9vQZWAbtmzZQp8+fVxun5KS4tIiS3vtlZ+ic59ozk5oRVrqCTZ/+wdrv9zN5uV/0GtoDPHDYjiwLdOrJebKt3elRF5NSYDsK9JTJb1CCCGE26ZMmUJqaqq3uyHqIaUUsd2iiO0WxdG92Wz+9g82Lttn3eREoS0aqHmJuB3rjrB6XipFZ1wrMedue1dIgOwLiovg+E4491Jv91Hhl4AAAAg3SURBVEQIIYSPCwwMJCBAfv2L6rXu1IxRd/Ui8/BpPvvHhtLgtETRGQs/fLydw3uyHJ5r+69HPNLekyXq5F+IL8jcA8VnZAZZCCGE2zZs2MCRI0ekDrJwStRZTaoEpyUKC4rZtSnd4TkKC2wv+nO3fXWl89wlAbIvKFmgJ5uECCGEcNPvv/8udZBFjVRXou7mv1/gsP0HT/7ssfaeImXefEF6KqCgpVSwEEIIIUTdqpcl6mrQ3hUyg+wLMlIgsgMEhXm7J0IIIYRoZOpziTpPkQDZF0gFCyGEEEJ4kadKzNW0vasl8mpKUizqu+JCOL5L8o+FEEIIIeqI0lp7uw+1QillAfLcOIU/4M7eir7ePgg448Xre7O9jJ3ce966dxrz2NWH9r48fjJ2vt3el8fP18cuVGvteIJYa90gHsBGN9u/28jbZ/h4/11uL2Pn9fY+O34ydj7f3mfHT8bO59v77Pj5+tg5+5AUizJfNfL2J718fW+2l7HzbntfHj8ZO99u78vjJ2Pn2+19efx8feyc0pBSLDZqrft7ux++SsbPdTJ27pHxc52MnXtk/FwnY+ceGT/X1dXYNaQZ5He93QEfJ+PnOhk798j4uU7Gzj0yfq6TsXOPjJ/r6mTsGswMshBCCCGEELWhIc0gCyGEEEII4TYJkIUQQgghhCinQQTISqnLlVLblVK7lFJPeLs/vkQptU8p9ZtSKlEptdHb/anvlFKzlFLpSqnkcseilFLfK6V2Wj9GerOP9Zmd8XtOKXXQeg8mKqWu8GYf6yulVKxSarVSKkUp9btS6gHrcbn/HKhm7OTec4JSKkQptV4ptdU6fs9bj3dSSq2z3nufKqWCvN3X+qaasZujlNpb7t5L8HZf6yullL9SaotSaqn16zq573w+QFZK+QNvAqOA7sBkpVR37/bK5wzTWifIilqnzAEur3TsCWCl1vpcYKX1a2HbHKqOH8Cr1nswQWu9rI775CuKgIe11t2AQcC91v/r5P5zzN7Ygdx7zigALtFa9wYSgMuVUoOAf2HG71zgBHCbF/tYX9kbO4BHy917id7rYr33AJBS7us6ue98PkAGBgC7tNZ7tNZngPnAWC/3STRQWus1QGalw2OBD6yffwCMq9NO+RA74yecoLU+rLXebP38FOYXRjvk/nOomrETTtBGjvXLQOtDA5cAn1uPy71nQzVjJ5yglIoBrgTes36tqKP7riEEyO2AA+W+TkP+46sJDXynlNqklLrT253xUa211ofB/CIGor3cH190n1IqyZqCISkCDiilOgJ9gHXI/VcjlcYO5N5zivVt7kQgHfge2A2c1FoXWV8iv3vtqDx2WuuSe+9F6733qlIq2ItdrM9mAI8BFuvXLaij+64hBMjKxjH568x5F2it+2JSVO5VSg3xdodEo/M20Bnz9uNh4BXvdqd+U0qFA18Af9FaZ3u7P77ExtjJveckrXWx1joBiMG8c9vN1svqtle+ofLYKaV6An8F4oDzgCjgcS92sV5SSl0FpGutN5U/bOOlHrnvGkKAnAbElvs6Bjjkpb74HK31IevHdGAh5j8+UTNHlVJnAVg/pnu5Pz5Fa33U+gvEAsxE7kG7lFKBmABvntb6S+thuf+cYGvs5N6rOa31SeAHTC53c6VUgPUp+d3rQLmxu9ya9qO11gXAbOTes+UCYIxSah8mffYSzIxyndx3DSFA3gCca13VGARcByzxcp98glKqiVKqacnnwEggufpWwoYlwM3Wz28GFnuxLz6nJLizGo/cgzZZc+/eB1K01tPLPSX3nwP2xk7uPecopVoppZpbPw8FLsXkca8GJlhfJveeDXbGLrXcH7UKk0Mr914lWuu/aq1jtNYdMbHdKq31FOrovmsQO+lZS/PMAPyBWVrrF73cJZ+glDobM2sMEAB8LGNXPaXUJ8BQoCVwFHgWWAQsANoD+4FrtdayEM0GO+M3FPMWtwb2AX8qyakVZZRSFwI/Ab9Rlo/3JCaXVu6/alQzdpORe88hpVQ8ZjGUP2ZibYHWepr1d8h8TIrAFuAG64yosKpm7FYBrTApA4nAXeUW84lKlFJDgUe01lfV1X3XIAJkIYQQQgghaktDSLEQQgghhBCi1kiALIQQQgghRDkSIAshhBBCCFGOBMhCCCGEEEKUIwGyEEIIIYQQ5UiALIQQ9ZxSqlgplVju8UQtnrujUkpqsAohRDkBjl8ihBDCy/KsW9UKIYSoAzKDLIQQPkoptU8p9S+l1Hrr4xzr8Q5KqZX/394dq2YRRGEYfj9CkECwURBBJE0qIYKIhaW3YBHFSlKlMVXIDeQKgmkiWAjWtqJYCKLYqZBW7BJIChEbETkW/wQGScpfd837wLJnzy7LTHc4zO4k+dTOl1v+QpJnST6242Z71UySR0l2k7xoO35J0qllgSxJwzf3xxKL5e7et6q6ATxksqMoLX5SVUvAU2Cr5beA11V1FbgG7Lb8IrBdVVeAr8DtKc9HkgbNnfQkaeCSfK+q+WPyX4BbVfU5ySywX1XnkhwCF6vqZ8vvVdX5JAfApX5b1iQLwMuqWmzXG8BsVW1Of2aSNEx2kCVp3OqE+KRnjvOji3/h9ymSTjkLZEkat+Xu/K7Fb4E7Lb4HvGnxK2AVIMlMkrN/a5CSNCZ2CSRp+OaSfOiun1fV0a/eziR5z6ThcbflHgCPk6wDB8D9ll8DdpKsMOkUrwJ7Ux+9JI2Ma5AlaaTaGuTrVXX4r8ciSf8Tl1hIkiRJHTvIkiRJUscOsiRJktSxQJYkSZI6FsiSJElSxwJZkiRJ6lggS5IkSZ3fO/Jl9ZPHT9AAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 720x180 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "log = plot_log_for(results, idx=44, loss=True, acc=False, figsize=(10.0, 2.5));\n",
    "plot_log_for(results, idx=34, loss=True, acc=False, figsize=(10.0, 2.5));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "annealers = {}\n",
    "epochs = 40\n",
    "\n",
    "#annealers['Linear'] = lambda x, lr=0.066: lr * (1 - x / epochs)\n",
    "\n",
    "annealers['Linear'] = lambda x, lr=0.066, end_lr=0.000, min_lr=0.005: \\\n",
    "    np.maximum(lr - (lr - end_lr) / epochs * x, min_lr)\n",
    "    \n",
    "annealers['ExpDecay'] = lambda x, lr=0.066, rate=0.94, min_lr=0.0001: \\\n",
    "    np.maximum(lr * rate**(x//2), min_lr)\n",
    "    \n",
    "#annealers['Custom'] = lambda x, start=0.04, end=0.005, center=24: \\\n",
    "#    (start - end) / (1 + np.e**((x - center) / 3)) + end\n",
    "    \n",
    "#annealers['ReduceOnPlateau'] = lambda x: reduce_on_plat(lr=0.04, factor=0.5, plats=[19, 27, 32], epochs=40)\n",
    "\n",
    "annealers['idx ' + str(log[0])] = lambda x: log[1]['lr']\n",
    "\n",
    "#annealers['Constant'] = lambda x: np.repeat(0.001, len(x))\n",
    "\n",
    "plot_annealers(annealers, epochs=epochs, figsize=(10.0, 5.0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Notes\n",
    "\n",
    "Best so far: 9957 - loss 0.0195, Adam with linear decay, 120 epochs, peak at roughly epoch 70, batch dropout(.2,.2,.2,.3)\n",
    "\n",
    "If data augs not listed in notes then they're:\n",
    "\n",
    "    rotation_range=15,\n",
    "    width_shift_range=0.1,\n",
    "    height_shift_range=0.1,\n",
    "    shear_range=15,\n",
    "    zoom_range=[0.9, 1]\n",
    "    \n",
    "- Good ReduceOnPlat points from Desmos: (0,0.04),(18,0.04),(19,0.02),(26,0.02),(27,0.01),(31,0.01),(32,0.005),(34,0.005)\n",
    "\n",
    "- 64 didn't increase accuracy over 32, in stem\n",
    "- second conv layer in stem doesn't increase accuracy\n",
    "- same' padding leads to overfitting with drop out (.2,.2,.2,.3), and validation accuracy stalls with (.3,.3,.3,.5) - also repeats in weights visualizations\n",
    "- 3,3 max pool hit .9952 with dropout (.2,.2,.2,.2)\n",
    "- no batch norm - less accuracy\n",
    "- label smoothing - loss < .02, val acc 0.5\n",
    "- he_uniform on all layers - vall_acc never got over .995 (got to .9948) , weight vis still looked like noise after 120 epochs\n",
    "- PRelu (alpha=0.25) - extreme overfitting, increase drop out?\n",
    "- he_uniform last layer, < 0.995\n",
    "- droput that preserves inception topology overfits (.2, .2, .2, .3)\n",
    "- label smoothing, dropout (.25,.25,.25,.25) overfits, < 0.995\n",
    "- label smoothing, combined dropout (.2,.2,.2,.2)(.4) overfits, < 0.995\n",
    "- dropout with noise shape seems to have weight vis that is less 'noisy' (might actually be from decreasing Adams learning rate too quickly)\n",
    "- RMSprop(lr=0.045, rho=0.9, epsilon=1.0, clipvalue=2.0),  # incep v3 paper setting max out at .9940, weight vis still noisy after 100 epochs\n",
    "- repeats in weight vis appears some what common with dropout only in last 2 modules\n",
    "\n",
    "Label Smoothing:\n",
    "- looks like it might stop validation loss from increasing when over training i.e. without label smoothing, there's a certain point where val loss stops decreasing and starts to trend upwards, with label smoothng there's no upward trend\n",
    "\n",
    "Things to try:\n",
    "- same padding with SGD\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**old notes, results now stored in results.csv**\n",
    "\n",
    "Best so far: 9957 - loss 0.0195, Adam with linear decay, 120 epochs, peak at roughly epoch 70, batch dropout(.2,.2,.2,.3)\n",
    "\n",
    "SGD 9952 - learning rate 0.0378 at epoch 20 (decayed from 0.066, momentum=0.95, schedule decay rate 0.94), epoch 24 train loss increaseses by 0.003 before that any train loss increases were by < 0.002\n",
    "\n",
    "if dropout isn't listed then it's (spatial dropout: 0,0,.3,.5)\n",
    "\n",
    "if data augs not listed then they're:\n",
    "\n",
    "    rotation_range=15,\n",
    "    width_shift_range=0.1,\n",
    "    height_shift_range=0.1,\n",
    "    shear_range=15,\n",
    "    zoom_range=[0.9, 1]\n",
    "\n",
    "### SGD\n",
    "- momentum = 0.97 hit 9943, lr 0.0330 at epoch 34 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.066)\n",
    "---\n",
    "- momentum = 0.96 hit 9950, lr 0.0165 at epoch 31 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.066)\n",
    "- momentum = 0.96 hit 9940, lr 0.0168 at epoch 30 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.067)\n",
    "- momentum = 0.96 hit 9936, lr 0.0165 at epoch 23 (reduce on plateau, factor=0.25, patience=3, initial lr = 0.066, no label smoothing)(spatial dropout: 0.2,0.2,0.2,0.3)\n",
    "- momentum = 0.96 hit 9926, lr 0.0082 at epoch 16 (reduce on plateau, factor=0.5, patience=0, initial lr = 0.066, no label smoothing)(spatial dropout: 0.2,0.2,0.3,0.5)\n",
    "- momentum = 0.96 hit 9945, lr 0.0041 at epoch 36 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.066)(spatial dropout: 0,0,.5,.5)\n",
    "- momentum = 0.96 hit 9945, lr 0.0297 at epoch 23 (schedule decay, rate=0.93, initial lr = 0.066)\n",
    "---\n",
    "- momentum = 0.95 hit 9943, lr 0.0165 at epoch 28 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.066)\n",
    "- momentum = 0.95 hit 9948 (loss=0.0171), lr 0.005 at epoch 34 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.04)\n",
    "- momentum = 0.95 hit 9945 (overfit), lr 0.0200 at epoch 27 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.04)(no rotation or skew)\n",
    "- momentum = 0.95 hit 9921, lr 0.0200 at epoch 31 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.04)(no rotation or skew)(spatial dropout: .5,.5,.5,.5)\n",
    "- momentum = 0.95 hit 9926, lr 0.0165 at epoch 33 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.066)(no rotation or skew)(spatial dropout: 0,.5,.5,.5)\n",
    "- momentum = 0.95 hit 9938, lr 0.0006 at epoch 29 (reduce on plateau, factor=0.5, patience=2, initial lr = 0.04)\n",
    "- momentum = 0.95 dud, stalled at 9926 (reduce on plateau, factor=0.5, patience=0, initial lr = 0.04)\n",
    "- momentum = 0.95 hit 9931, lr 0.0181 at epoch 20 (schedule decay, rate=0.92, initial lr = 0.04)\n",
    "- momentum = 0.95 hit 9952, lr 0.0378 at epoch 20 (schedule decay, rate=0.94, initial lr = 0.066)\n",
    "- momentum = 0.95 hit 9945, lr 0.0413 at epoch 16 (linear decay, epochs=40, initial lr = 0.066)\n",
    "- momentum = 0.95 hit 9945, lr 0.0165 at epoch 28 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.066, no label smoothing)\n",
    "- momentum = 0.95 hit 9943, lr 0.0231 at epoch 35 (schedule decay, rate=0.94, initial lr = 0.066)(he_uniform on last 2 modules)\n",
    "---\n",
    "- momentum = 0.94 hit 9945, lr 0.0165 at epoch 25 (reduce on plateau, factor=0.5, patience=3, initial lr = 0.066)\n",
    "\n",
    "### Adamax\n",
    "- beta_1 = 0.5 hit 9950, lr 0.001 at epoch 32 (reduce on plateau, factor=0.5, patience=0, initial lr = 0.03, min_lr = 0.001)\n",
    "- beta_1 = 0.5 hit 9945 (no overfit), lr 0.005 at epoch 30 (reduce on plateau, factor=0.9, patience=0, initial lr = 0.03, min_lr = 0.0005)(spatial dropout: .0,.0,.5,.5)\n",
    "- beta_1 = 0.49 hit 9948, at epoch 34 (linear decay, initial lr = 0.01, 40 epochs, min_lr = 0.0002)\n",
    "- beta_1 = 0.49 hit 9945 (overfit), at epoch 37 (linear decay, initial lr = 0.02, 40 epochs, min_lr = 0.0002)(batch dropout: .2,.2,.2,.3)\n",
    "- beta_1 = 0.49 hit 9950 (overfit), lr 0.001 at epoch 18 (reduce on plateau, monitor='loss', factor=0.5, patience=0, initial lr = 0.04, min_lr = 0.0002)\n",
    "- beta_1 = 0.49, beta_2 = 0.95, hit 9933, lr 0.005 at epoch 30 (reduce on plateau, monitor='val_loss', factor=0.5, patience=3, initial lr = 0.04, min_lr = 0.0002)\n",
    "- beta_1 = 0.49 hit 9938, lr 0.01 at epoch 19 (reduce on plateau, monitor='loss', factor=0.5, patience=0, initial lr = 0.04, min_lr = 0.0002)(batch dropout: .5,.5,.5,.5)"
   ]
  }
 ],
 "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": "3.6.6"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "state": {},
    "version_major": 2,
    "version_minor": 0
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}