homework/Day_089_HW.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Work\n",
    "1. 請自行定義一個 loss function, 為 0.3 * focal loss + 0.7 cross-entropy，訓練並比較結果\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import keras\n",
    "\n",
    "# Disable GPU\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train, test = keras.datasets.cifar10.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## 資料前處理\n",
    "def preproc_x(x, flatten=True):\n",
    "    x = x / 255.\n",
    "    if flatten:\n",
    "        x = x.reshape((len(x), -1))\n",
    "    return x\n",
    "\n",
    "def preproc_y(y, num_classes=10):\n",
    "    if y.shape[-1] == 1:\n",
    "        y = keras.utils.to_categorical(y, num_classes)\n",
    "    return y    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train, y_train = train\n",
    "x_test, y_test = test\n",
    "\n",
    "# Preproc the inputs\n",
    "x_train = preproc_x(x_train)\n",
    "x_test = preproc_x(x_test)\n",
    "\n",
    "# Preprc the outputs\n",
    "y_train = preproc_y(y_train)\n",
    "y_test = preproc_y(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import BatchNormalization\n",
    "def build_mlp(input_shape, output_units=10, num_neurons=[512, 256, 128]):\n",
    "    \"\"\"Code Here\n",
    "    建立你的神經網路\n",
    "    \"\"\"\n",
    "    input_layer  = keras.layers.Input(input_shape)\n",
    "    \n",
    "    for i , num in enumerate(num_neurons):\n",
    "        if i ==0:\n",
    "            x = keras.layers.Dense(units = num,activation='relu',name='HIDDEN_LAYER_'+str(i+1))(input_layer)\n",
    "            x = BatchNormalization()(x)\n",
    "            x  = Dropout(0.1)(x)\n",
    "        else:\n",
    "            x = keras.layers.Dense(units = num,activation='relu',name='HIDDEN_LAYER_'+str(i+1))(x)\n",
    "            x = BatchNormalization()(x)\n",
    "            x  = Dropout(0.1)(x)\n",
    "            \n",
    "    out = keras.layers.Dense(units=output_units,activation='softmax',name=\"OUTPUT_LAYER\")(x)   \n",
    "    \n",
    "    model = keras.models.Model(inputs=[input_layer],outputs=[out])\n",
    "    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## 超參數設定\n",
    "LEARNING_RATE = 1e-3\n",
    "EPOCHS = 25\n",
    "BATCH_SIZE = 1024\n",
    "MOMENTUM = 0.95"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import keras.backend as K\n",
    "\n",
    "\"\"\"Code Here\n",
    "撰寫一個 loss function, 使其可以結合 focal loss 與 crossentropy loss\n",
    "\"\"\"\n",
    "def combined_loss():\n",
    "    \"\"\"Define the customized loss.\"\"\"\n",
    "    gamma = float(gamma)\n",
    "    alpha = float(alpha)\n",
    "    \n",
    "    def focal_loss_fixed(y_true, y_pred):\n",
    "        \"\"\"Focal loss for multi-classification\n",
    "        FL(p_t)=-alpha(1-p_t)^{gamma}ln(p_t)\n",
    "        \"\"\"\n",
    "        epsilon = 1e-8\n",
    "        y_true = tf.convert_to_tensor(y_true, tf.float32)\n",
    "        y_pred = tf.convert_to_tensor(y_pred, tf.float32)\n",
    "\n",
    "        model_out = tf.add(y_pred, epsilon)\n",
    "        ce = tf.multiply(y_true, -tf.log(model_out))\n",
    "        weight = tf.multiply(y_true, tf.pow(tf.subtract(1., model_out), gamma))\n",
    "        fl = tf.multiply(alpha, tf.multiply(weight, ce))\n",
    "        reduced_fl = tf.reduce_max(fl, axis=1)\n",
    "        \n",
    "        return ce_weight*tf.reduce_mean(reduced_fl) + (1-ce_weight) * (-tf.log(model_out))\n",
    "\n",
    "    return focal_loss_fixed\n",
    "\n",
    "ce_weights_list = [0., 0.3, 0.5, 0.7, 1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import itertools\n",
    "results = {}\n",
    "\n",
    "for i, ce_w in enumerate(ce_weights_list):\n",
    "    print(\"Numbers of exp: %i, ce_weight: %.2f\" % (i, ce_w))\n",
    "\n",
    "    model = build_mlp(input_shape=x_train.shape[1:])\n",
    "    model.summary()\n",
    "    optimizer = keras.optimizers.SGD(lr=LEARNING_RATE, nesterov=True, momentum=MOMENTUM)\n",
    "    \"\"\"Code Here\n",
    "    將自定義的 loss function 加入模型\n",
    "    \"\"\"\n",
    "    model.compile(loss=combined_loss(), metrics=[\"accuracy\"], optimizer=optimizer)\n",
    "    \n",
    "    model.fit(x_train, y_train, \n",
    "              epochs=EPOCHS, \n",
    "              batch_size=BATCH_SIZE, \n",
    "              validation_data=(x_test, y_test), \n",
    "              shuffle=True\n",
    "             )\n",
    "\n",
    "    # Collect results\n",
    "    exp_name_tag = (\"exp-%s\" % (i))\n",
    "    results[exp_name_tag] = {'train-loss': model.history.history[\"loss\"],\n",
    "                             'valid-loss': model.history.history[\"val_loss\"],\n",
    "                             'train-acc': model.history.history[\"acc\"],\n",
    "                             'valid-acc': model.history.history[\"val_acc\"]}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.cm as mplcm\n",
    "import matplotlib.colors as colors\n",
    "%matplotlib inline\n",
    "NUM_COLORS = len(results.keys())\n",
    "\n",
    "cm = plt.get_cmap('gist_rainbow')\n",
    "cNorm  = colors.Normalize(vmin=0, vmax=NUM_COLORS-1)\n",
    "scalarMap = mplcm.ScalarMappable(norm=cNorm, cmap=cm)\n",
    "color_bar = [scalarMap.to_rgba(i) for i in range(NUM_COLORS)]\n",
    "\n",
    "plt.figure(figsize=(8,6))\n",
    "for i, cond in enumerate(results.keys()):\n",
    "    plt.plot(range(len(results[cond]['train-loss'])),results[cond]['train-loss'], '-', label=cond, color=color_bar[i])\n",
    "    plt.plot(range(len(results[cond]['valid-loss'])),results[cond]['valid-loss'], '--', label=cond, color=color_bar[i])\n",
    "plt.title(\"Loss\")\n",
    "plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\n",
    "plt.show()\n",
    "\n",
    "plt.figure(figsize=(8,6))\n",
    "for i, cond in enumerate(results.keys()):\n",
    "    plt.plot(range(len(results[cond]['train-acc'])),results[cond]['train-acc'], '-', label=cond, color=color_bar[i])\n",
    "    plt.plot(range(len(results[cond]['valid-acc'])),results[cond]['valid-acc'], '--', label=cond, color=color_bar[i])\n",
    "plt.title(\"Accuracy\")\n",
    "plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\n",
    "plt.show()"
   ]
  }
 ],
 "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": "3.6.9"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "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()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}