ComparisonIncreasingDropout.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T17:37:48.238493Z",
     "start_time": "2019-01-05T17:37:48.223469Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "# initialization\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "from experiment_train import *\n",
    "from experiment_datasets import *\n",
    "from helpers import *"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## See steps in MNIST notebook"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T18:02:43.090987Z",
     "start_time": "2019-01-05T18:02:43.076776Z"
    }
   },
   "outputs": [],
   "source": [
    "# network configuration\n",
    "N = [100, 50, 50, 30]\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",
    "# prob. of failure at inference\n",
    "p_inference = get_p_arr(0.1)\n",
    "\n",
    "# prob. of failure at train\n",
    "p_trains = np.linspace(0, 0.5, 5)\n",
    "\n",
    "# Lips. coeff\n",
    "KLips = 1.\n",
    "\n",
    "# activation function\n",
    "activation = 'sigmoid'\n",
    "\n",
    "# number of repeats for error estimation\n",
    "# see below for explanation\n",
    "repetitions = 10\n",
    "\n",
    "# number of runs for each p\n",
    "runs = 20\n",
    "\n",
    "def get_dataset():\n",
    "    (x_train, y_train), (x_test, y_test) = boston_housing.load_data()\n",
    "    return x_train, y_train, x_test, y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T18:11:18.546713Z",
     "start_time": "2019-01-05T18:11:18.515886Z"
    }
   },
   "outputs": [],
   "source": [
    "def get_results(p):\n",
    "    # obtaining the distribution after 10 epochs\n",
    "    results = []\n",
    "    \n",
    "    # current prob. of failure array\n",
    "    p_train = get_p_arr(p)\n",
    "    \n",
    "    def unravel(name, data):\n",
    "        return {name + '_train': data['train'], name + '_test': data['test']}\n",
    "    \n",
    "    for _ in tqdm(range(runs)):\n",
    "        exp = TrainExperiment(*get_dataset(), N = N, p_inference = p_inference, p_train = p_train, KLips = KLips, epochs = 10,\n",
    "                                      activation = activation, reg_type = None, reg_coeff = 0, do_print = False,\n",
    "                                      name = 'experiment_dropout', seed = None, batch_size = 100)\n",
    "        \n",
    "        results += [{**unravel('correct_mae', exp.get_mae_correct()),\n",
    "                     **unravel('crashing_mae', exp.get_mae_crash(repetitions = repetitions)),\n",
    "                     'v2_var': np.mean(exp.get_bound_v2(exp.x_train)['std'] ** 2),\n",
    "                     'v2_mean': np.mean(exp.get_bound_v2(exp.x_train)['mean']),\n",
    "                     \n",
    "                     'v3_var': np.mean(exp.get_bound_v3(exp.x_train)['std'] ** 2),\n",
    "                     'v3_mean': np.mean(exp.get_bound_v3(exp.x_train)['mean']),\n",
    "                               \n",
    "                     'v4_var': np.mean(exp.get_bound_v4(exp.x_train)['std'] ** 2),\n",
    "                     'v4_mean': np.mean(exp.get_bound_v4(exp.x_train)['mean']),\n",
    "                    }]\n",
    "        K.clear_session()\n",
    "        \n",
    "    # list of all keys\n",
    "    keys = list(results[0].keys())\n",
    "    \n",
    "    # return results\n",
    "    return {key: [value[key] for value in results] for key in keys}\n",
    "\n",
    "def process_results_single(results):\n",
    "    for key, value in results.items():\n",
    "        plt.figure()\n",
    "        plt.title('%s: %.3f +- %.3f med %.3f' % (key, np.mean(value), np.std(value), np.median(value)))\n",
    "        plt.hist(value)\n",
    "        plt.xlabel(key)\n",
    "        plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "deletable": false,
    "editable": false,
    "run_control": {
     "frozen": true
    }
   },
   "outputs": [],
   "source": [
    "exp = TrainExperiment(*get_dataset(), N = N, p_inference = p_inference, p_train = get_p_arr(0.01), KLips = KLips, epochs = 10,\n",
    "                                      activation = activation, reg_type = None, reg_coeff = 0, do_print = False,\n",
    "                                      name = 'experiment_dropout', seed = None, batch_size = 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "deletable": false,
    "editable": false,
    "run_control": {
     "frozen": true
    }
   },
   "outputs": [],
   "source": [
    "# array of repetitions values\n",
    "rs = np.arange(30)[1:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T17:34:08.654323Z",
     "start_time": "2019-01-05T17:34:08.633327Z"
    },
    "deletable": false,
    "editable": false,
    "run_control": {
     "frozen": true
    }
   },
   "outputs": [],
   "source": [
    "# obtaining accuracy for each repetition parameter\n",
    "results = [exp.get_mae_crash(repetitions = r) for r in tqdm(rs)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "deletable": false,
    "editable": false,
    "run_control": {
     "frozen": true
    },
    "scrolled": true
   },
   "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"
    }
   ],
   "source": [
    "plt.figure()\n",
    "plt.title('Crash MAE and repetitions')\n",
    "plt.plot([x['train'] for x in results], label = 'train')\n",
    "plt.plot([x['test'] for x in results], label = 'test')\n",
    "plt.plot(range(len(results)), [exp.get_mae_correct()['train']] * len(results), label = 'train_correct')\n",
    "plt.plot(range(len(results)), [exp.get_mae_correct()['test']] * len(results), label = 'test_correct')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T17:38:33.658427Z",
     "start_time": "2019-01-05T17:38:33.642392Z"
    },
    "deletable": false,
    "editable": false,
    "run_control": {
     "frozen": true
    }
   },
   "outputs": [],
   "source": [
    "# 10 repetitions is enough"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T18:25:25.386970Z",
     "start_time": "2019-01-05T18:11:21.396036Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [02:30<00:00,  6.82s/it]\n",
      "100%|██████████| 20/20 [02:12<00:00,  6.65s/it]\n",
      "100%|██████████| 20/20 [02:12<00:00,  6.62s/it]\n",
      "100%|██████████| 20/20 [04:39<00:00, 13.70s/it]\n",
      "100%|██████████| 20/20 [02:29<00:00,  6.66s/it]\n"
     ]
    }
   ],
   "source": [
    "# obtaining results\n",
    "results_array = [get_results(p) for p in p_trains]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T18:30:34.885382Z",
     "start_time": "2019-01-05T18:30:34.775430Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f9f857ccb70>"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "plt.figure()\n",
    "\n",
    "def plot_quantity(results_array, name):\n",
    "    means = np.array([np.mean(r[name]) for r in results_array])\n",
    "    stds  = [np.std(r[name]) for r in results_array]\n",
    "    plt.plot(p_trains, means, label = name)\n",
    "    plt.fill_between(p_trains, means - stds, means + stds, alpha = 0.5)\n",
    "\n",
    "plot_quantity(results_array, 'correct_mae_train')\n",
    "plot_quantity(results_array, 'crashing_mae_train')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T17:48:30.511462Z",
     "start_time": "2019-01-05T17:48:30.423001Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f9ff01e4cf8>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "plt.figure()\n",
    "plt.plot(p_trains, [np.mean(r['v4_var']) for r in results_array], label = 'v4_var')\n",
    "#plt.plot(p_trains, [np.mean(np.abs(r['v4_mean'])) for r in results_array], label = 'v4_mean')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T17:48:33.889262Z",
     "start_time": "2019-01-05T17:48:33.790309Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f9f94449a90>"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "plt.figure()\n",
    "plt.plot(p_trains, [np.mean(r['v3_var']) for r in results_array], label = 'v3_var')\n",
    "#plt.plot(p_trains, [np.mean(np.abs(r['v3_mean'])) for r in results_array], label = 'v3_mean')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-05T17:48:35.534350Z",
     "start_time": "2019-01-05T17:48:35.436430Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f9ff055b630>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "plt.figure()\n",
    "plt.plot(p_trains, [np.mean(r['v2_var']) for r in results_array], label = 'v2_var')\n",
    "#plt.plot(p_trains, [np.mean(np.abs(r['v2_mean'])) for r in results_array], label = 'v2_mean')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.6"
  },
  "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": "515px",
    "left": "1484px",
    "right": "20px",
    "top": "120px",
    "width": "438px"
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}