ゼロから深層学習.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 自分で深層学習を実装してみる！！"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 3桁の２進数を、１０進数に変換するネット"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.2 0.5 0.8 1.  1.2 0.7 0.9 1.1]\n",
      "[1.36021119 3.10758575 3.24562428 3.59474913 4.24387399]\n",
      "[0.95534701 1.99027342 2.10238961 3.50477923 3.03731602 3.50477923\n",
      " 3.97224243 3.97224243]\n",
      "[0.01227909 0.03456441 0.03866523 0.15717038 0.09848135 0.15717038\n",
      " 0.25083457 0.25083457]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "def sigmoid(x):\n",
    "    return 1 / (1 + np.exp(-x))\n",
    "\n",
    "def identity_function(x):#恒等関数\n",
    "    return x \n",
    "\n",
    "def softmax(a):\n",
    "    c = np.max(a)\n",
    "    exp_a = np.exp(a - c)#最大値で引くことで、オーバーフローを防ぐ\n",
    "    sum_exp_a = np.sum(exp_a)\n",
    "    y = exp_a /sum_exp_a\n",
    "    return y\n",
    "\n",
    "\n",
    "def change_one_hot(lst):\n",
    "    # 最大値を取得\n",
    "    max_value = np.max(lst)\n",
    "\n",
    "    # 最大値だけ1、残りを0の配列に置き換える\n",
    "    result = np.zeros(8)\n",
    "    result[np.argmax(lst)] = 1\n",
    "\n",
    "    return result\n",
    "\n",
    "def sum_squared_error(y, t):#２乗和誤差\n",
    "    return 0.5 * np.sum((y-t)**2)\n",
    "\n",
    "def bibun_numerical_gradient(f, x):\n",
    "    h = 1e-4#0.0001\n",
    "    return (f(x+h)-f(x-h))/(2*h)\n",
    "\n",
    "\n",
    "\n",
    "def init_network(): #3層のネットワークを定義\n",
    "    network = {}\n",
    "    network['W1'] = np.array([[0.1, 0.3, 0.5, 0.6, 0.7, 0.1, 0.2, 0.3], \n",
    "                              [0.2, 0.4, 0.6, 0.1, 0.3, 0.5, 0.6, 0.2], \n",
    "                              [0.3, 0.1, 0.2, 0.4, 0.2, 0.5, 0.1, 0.4], \n",
    "                              ]) #3つのニューロンから、8つのニューロンに出力する際の重み\n",
    "    network['b1'] = np.array([0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8,]) #バイアスを8つのニューロンにたす\n",
    "\n",
    "\n",
    "    network['W2'] = np.array([[0.1, 0.4, 0.5, 0.6, 0.7],\n",
    "                              [0.2, 0.5, 0.5, 0.6, 0.7],\n",
    "                              [0.3, 0.6, 0.5, 0.6, 0.7],\n",
    "                              [0.2, 0.5, 0.5, 0.6, 0.7],\n",
    "                              [0.3, 0.6, 0.5, 0.6, 0.7],\n",
    "                              [0.2, 0.5, 0.5, 0.6, 0.7],\n",
    "                              [0.3, 0.6, 0.5, 0.6, 0.7],\n",
    "                              [0.2, 0.5, 0.5, 0.6, 0.7]]) #8つのニューロンから5つのニューロンに出力する際の重み\n",
    "    \n",
    "    network['b2'] = np.array([0.1, 0.2, 0.5, 0.3, 0.4])#5つのニューロンにたすバイアス\n",
    "    network['W3'] = np.array([[0.1, 0.3, 0.3, 0.6, 0.5, 0.6, 0.7, 0.7],\n",
    "                              [0.2, 0.4, 0.3, 0.6, 0.5, 0.6, 0.7, 0.7],\n",
    "                              [0.2, 0.4, 0.3, 0.6, 0.5, 0.6, 0.7, 0.7],\n",
    "                              [0.2, 0.4, 0.3, 0.6, 0.5, 0.6, 0.7, 0.7],\n",
    "                              [0.2, 0.4, 0.3, 0.6, 0.5, 0.6, 0.7, 0.7]]) #5つのニューロンから8つのニューロンに出力する際の重み\n",
    "    network['b3'] = np.array([0.1, 0.2, 0.7, 0.7, 0.7, 0.7, 0.7, 0.7]) #バイアスを8つのニューロンにたす\n",
    "\n",
    "    return network\n",
    "\n",
    "def forward(network, x):\n",
    "    W1, W2, W3 = network['W1'], network['W2'], network['W3']\n",
    "    b1, b2, b3 = network['b1'], network['b2'], network['b3']\n",
    "\n",
    "    a1 = np.dot(x, W1) + b1\n",
    "    print(a1)\n",
    "    z1 = sigmoid(a1)\n",
    "    a2 = np.dot(z1, W2) + b2\n",
    "    print(a2)\n",
    "    z2 = sigmoid(a2)\n",
    "    a3 = np.dot(z2, W3) + b3\n",
    "    print(a3)\n",
    "    y = softmax (a3)\n",
    "\n",
    "    return y\n",
    "\n",
    "# network = init_network()\n",
    "# x = np.array([1, 0, 0])\n",
    "# y = forward(network, x)\n",
    "# print(y)\n",
    "# error = sum_squared_error(y, t)\n",
    "\n",
    "for i in range(10):\n",
    "    x = np.array([[1, 0, 0]])#入力\n",
    "    t = np.array([[0, 0, 0, 0, 1, 0, 0, 0]])\n",
    "    y = forward(network, x)\n",
    "    print(y)\n",
    "    error = sum_squared_error(y, t)\n",
    "    for i in range(W1.size):\n",
    "        koubai = bibun_numerical_gradient(f, x)\n",
    "\n",
    "        dw1 = np.append(dw1,np.append[koubai], axis(0))\n",
    "\n",
    "        bibun_numerical_gradient(error) #傾き求める\n",
    "\n",
    "    \n",
    "\n",
    "\n",
    "\n",
    "\n",
    "\n",

    "    # 微分して求めた式\n",
    "    W1 -= learning_rate * dL_dW1\n",
    "    W2 -= learning_rate * dL_dW2\n",
    "    W3 -= learning_rate * dL_dW3\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "\n",
    "def change_one_hot(lst):\n",
    "    # 最大値を取得\n",
    "    max_value = np.max(lst)\n",
    "\n",
    "    # 最大値だけ1、残りを0の配列に置き換える\n",
    "    result = np.zeros(8)\n",
    "    result[np.argmax(lst)] = 1\n",
    "\n",
    "    return result\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "takara",
   "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": "3.10.9"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}