2.NLP_PREP/2.1.1.tf.keras.layers.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.1.1 tf.keras.layers"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 라이브러리 불러오기 및 상수값 설정"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "INPUT_SIZE = (20, 1)\n",
    "CONV_INPUT_SIZE = (1, 28, 28)\n",
    "IS_TRAINING = True"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dense Layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "input = tf.placeholder(tf.float32, shape = INPUT_SIZE)\n",
    "output = tf.keras.layers.Dense(units = 10, activation = tf.nn.sigmoid)(input) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dense Layer with 1 hidden layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "input = tf.placeholder(tf.float32, shape = INPUT_SIZE)\n",
    "hidden = tf.keras.layers.Dense(units = 10, activation = tf.nn.sigmoid)(input)\n",
    "output = tf.keras.layers.Dense(units = 2, activation = tf.nn.sigmoid)(hidden)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "input = tf.placeholder(tf.float32, shape = INPUT_SIZE)\n",
    "dropout = tf.keras.layers.Dropout(rate = 0.5)(input)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dense Layer with 1 hidden layer and dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "input = tf.placeholder(tf.float32, shape = INPUT_SIZE)\n",
    "dropout = tf.keras.layers.Dropout(rate = 0.2)(input)\n",
    "hidden = tf.keras.layers.Dense(units = 10, activation = tf.nn.sigmoid)(dropout)\n",
    "output = tf.keras.layers.Dense(units = 2, activation = tf.nn.sigmoid)(hidden)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Convolutional layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "input = tf.placeholder(tf.float32, shape = CONV_INPUT_SIZE)\n",
    "conv = tf.keras.layers.Conv1D(\n",
    "         filters=10,\n",
    "         kernel_size=3,\n",
    "         padding='same',\n",
    "         activation=tf.nn.relu)(input)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Convolutional layer with dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "input = tf.placeholder(tf.float32, shape = CONV_INPUT_SIZE)\n",
    "dropout = tf.keras.layers.Dropout(rate=0.2)(input)\n",
    "conv = tf.keras.layers.Conv1D(\n",
    "         filters=10,\n",
    "         kernel_size=3,\n",
    "         padding='same',\n",
    "         activation=tf.nn.relu)(dropout)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Input -> Dropout -> Convolutional layer -> MaxPooling -> Dense layer with 1 hidden layer -> Output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "input = tf.placeholder(tf.float32, shape = CONV_INPUT_SIZE)\n",
    "dropout = tf.keras.layers.Dropout(rate = 0.2)(input)\n",
    "conv = tf.keras.layers.Conv1D(\n",
    "         filters=10,\n",
    "         kernel_size=3,\n",
    "         padding='same',\n",
    "         activation=tf.nn.relu)(dropout)\n",
    "max_pool = tf.keras.layers.MaxPool1D(pool_size = 3, padding = 'same')(conv)\n",
    "flatten = tf.keras.layers.Flatten()(max_pool)\n",
    "hidden = tf.keras.layers.Dense(units = 50, activation = tf.nn.relu)(flatten)\n",
    "output = tf.keras.layers.Dense(units = 10, activation = tf.nn.softmax)(hidden)  "
   ]
  },
  {
   "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}