hw1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "c5d7f5f5",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import warnings\n",
    "\n",
    "from sklearn.datasets import load_iris\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "warnings.simplefilter(\"ignore\")\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "5ff0046f",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Model:\n",
    "    def __init__(self):\n",
    "        self.layers = []\n",
    "        self.input_layer = None\n",
    "        self.loss = None\n",
    "\n",
    "    def add(self, layer):\n",
    "        self.layers.append(layer)\n",
    "\n",
    "    def set(self, loss):\n",
    "        self.loss = loss\n",
    "\n",
    "    def fit(self, X, y, epochs=1, learning_rate=0.01, print_every=1):\n",
    "        \"\"\"\n",
    "        Тренируем нашу сеть на выборке Х, y\n",
    "        :param epochs: Количество эпох\n",
    "        :param print_every: Как часто печатаем лог\n",
    "        :param learning_rate - шаг коррекции весов\n",
    "        :return:\n",
    "        \"\"\"\n",
    "        errors = []\n",
    "        for epoch in range(1, epochs + 1):\n",
    "            output = self.forward(X)\n",
    "\n",
    "            # Костыль для backpropagation\n",
    "            # TODO сделать через backward\n",
    "            error = y - output\n",
    "            errors.append(np.mean(np.abs(error)))\n",
    "            for layer in self.layers[::-1]:\n",
    "                # если текущий слой - активационный\n",
    "                if hasattr(layer, 'derive'):\n",
    "                    layer_delta = error * layer.derive(layer.output)\n",
    "                    error = layer_delta.dot(layer.prev.weights.T)\n",
    "                    # обновляем веса\n",
    "                    layer.prev.weights += layer.prev.input.T.dot(layer_delta) * learning_rate\n",
    "\n",
    "                else:\n",
    "                    continue\n",
    "\n",
    "            if epoch % print_every == 0:\n",
    "                print(f\"Loss on epoch {epoch}: {self.loss.calculate(output, y)}\")\n",
    "        return errors\n",
    "\n",
    "    def predict_proba(self, X):\n",
    "        return self.forward(X)\n",
    "\n",
    "    def predict(self, X):\n",
    "        return np.round(self.forward(X))\n",
    "\n",
    "    def finalize(self):\n",
    "        self.input_layer = LayerInput()\n",
    "        layer_count = len(self.layers)\n",
    "\n",
    "        # Проходимся по слоям и для каждого создаем атрибуты prev, next указывающие\n",
    "        # на предыдущий и следующий слой соответственно\n",
    "        for i in range(layer_count):\n",
    "            if i == 0:\n",
    "                # Если первый слой - то для него предыдущим будет входной\n",
    "                self.layers[i].prev = self.input_layer\n",
    "                self.layers[i].next = self.layers[i + 1]\n",
    "            elif i < layer_count - 1:\n",
    "                self.layers[i].prev = self.layers[i - 1]\n",
    "                self.layers[i].next = self.layers[i + 1]\n",
    "            else:\n",
    "                # Здесь замыкающим будет Loss функция\n",
    "                self.layers[i].prev = self.layers[i - 1]\n",
    "                self.layers[i].next = self.loss\n",
    "\n",
    "    def forward(self, X):\n",
    "        # TODO Проверка на существование input_layer\n",
    "        self.input_layer.forward(X)\n",
    "\n",
    "        for layer in self.layers:\n",
    "            layer.forward(layer.prev.output)\n",
    "\n",
    "        return layer.output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "5b2bb735",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LayerInput:\n",
    "    \"\"\"\n",
    "    Dummy класс для создания единообразия с прочими слоями. Содержит входные значения\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(self):\n",
    "        self.output = None\n",
    "\n",
    "    def forward(self, inputs):\n",
    "        self.output = inputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "1759c61f",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LayerDense:\n",
    "    \"\"\"\n",
    "    Класс полносвязанной слоя. В init'e инициализируем веса рандомно в нормальном\n",
    "    распределении со стандартным отклонением sigma. Bias инициализируем нулями\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(self, n_inputs, n_neurons, sigma=0.1):\n",
    "        self.weights = sigma * np.random.randn(n_inputs, n_neurons)\n",
    "        # self.biases = np.zeros((1, n_neurons))\n",
    "        self.output = None\n",
    "        self.input = None\n",
    "\n",
    "    def forward(self, inputs):\n",
    "        \"\"\"\n",
    "        Шаг прямого распространения - скалярно перемножаем веса на входные значения плюс смещение.\n",
    "        \"\"\"\n",
    "        self.input = inputs\n",
    "        self.output = np.dot(inputs, self.weights) #+ self.biases"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "885a601b",
   "metadata": {},
   "outputs": [],
   "source": [
    "class ActivationReLU:\n",
    "    \"\"\"\n",
    "    Функция активации ReLU. Выглядит как f(x) = max(0, x)\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(self):\n",
    "        self.output = None\n",
    "\n",
    "    def forward(self, inputs):\n",
    "        self.output = np.maximum(0, inputs)\n",
    "\n",
    "    @staticmethod\n",
    "    def derive(inputs):\n",
    "        \"\"\"\n",
    "        \"Производная\" функции ReLU.\n",
    "        \"\"\"\n",
    "        return np.where(inputs > 0, 1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "f08addef",
   "metadata": {},
   "outputs": [],
   "source": [
    "class ActivationSigmoid:\n",
    "    \"\"\"\n",
    "    Функция активации Sigmoid. Выглядит как f(x) = 1 / (1 + exp(-x))\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(self):\n",
    "        self.output = None\n",
    "\n",
    "    def forward(self, inputs):\n",
    "        self.output = 1 / (1 + np.exp(-inputs))\n",
    "\n",
    "    @staticmethod\n",
    "    def derive(inputs):\n",
    "        \"\"\"\n",
    "        Производная Sigmoid\n",
    "        \"\"\"\n",
    "        return inputs * (1 - inputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "7fc64966",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LossMeanSquaredError:\n",
    "    def calculate(self, output, y):\n",
    "        sample_losses = self.forward(output, y)\n",
    "        return np.mean(sample_losses)\n",
    "\n",
    "    @staticmethod\n",
    "    def forward(y_pred, y_true):\n",
    "        sample_losses = np.mean((y_true - y_pred) ** 2, axis=-1)\n",
    "        return sample_losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "49fc37ea",
   "metadata": {},
   "outputs": [],
   "source": [
    "def to_one_hot(input_y):\n",
    "    n_col = np.amax(input_y) + 1\n",
    "    binarized = np.zeros((len(input_y), n_col))\n",
    "    for i in range(len(input_y)):\n",
    "        binarized[i, input_y[i]] = 1.\n",
    "    return binarized\n",
    "\n",
    "\n",
    "def normalize(X, axis=-1, order=2):\n",
    "    l2 = np.atleast_1d(np.linalg.norm(X, order, axis))\n",
    "    l2[l2 == 0] = 1\n",
    "    return X / np.expand_dims(l2, axis)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "9826bf63",
   "metadata": {},
   "outputs": [],
   "source": [
    "iris = load_iris()\n",
    "iris_df = pd.DataFrame(iris.data, columns=iris.feature_names)\n",
    "target_df = iris.target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "cadcd3af",
   "metadata": {},
   "outputs": [],
   "source": [
    "x = normalize(iris_df.values)\n",
    "y = to_one_hot(target_df.flatten())\n",
    "X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.33)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "b33cd4e1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss on epoch 25000: 0.013260664106547524\n",
      "Loss on epoch 25000: 0.013548583994203656\n",
      "Loss on epoch 25000: 0.012563725605526939\n",
      "Loss on epoch 25000: 0.013430656309074982\n",
      "Loss on epoch 25000: 0.013335438506993404\n",
      "Loss on epoch 25000: 0.013203681668713487\n",
      "Loss on epoch 25000: 0.013063958277622054\n",
      "Loss on epoch 25000: 0.013037753311579683\n",
      "Loss on epoch 25000: 0.01342205234821403\n",
      "Loss on epoch 25000: 0.01319113222322532\n"
     ]
    }
   ],
   "source": [
    "# TODO сделать функцию обнуления весов\n",
    "batch_errors = []\n",
    "for _ in range(10):\n",
    "    nn_model = Model()\n",
    "    nn_model.add(LayerDense(4, 5))\n",
    "    nn_model.add(ActivationSigmoid())\n",
    "    nn_model.add(LayerDense(5, 4))\n",
    "    nn_model.add(ActivationSigmoid())\n",
    "    nn_model.add(LayerDense(4, 3))\n",
    "    nn_model.add(ActivationSigmoid())\n",
    "    nn_model.set(\n",
    "        loss=LossMeanSquaredError()\n",
    "    )\n",
    "    nn_model.finalize()\n",
    "\n",
    "    errors = nn_model.fit(X_train, y_train, learning_rate=0.01, epochs=25000, print_every=25000)\n",
    "    batch_errors.append(errors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "b36dea31",
   "metadata": {},
   "outputs": [],
   "source": [
    "colors = [\n",
    "    \"b\",\n",
    "    \"g\",\n",
    "    \"r\",\n",
    "    \"c\",\n",
    "    \"m\",\n",
    "    \"y\",\n",
    "    \"k\",\n",
    "    \"orange\",\n",
    "    \"pink\",\n",
    "    \"brown\"\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "835f5593",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Аккуратность нейронной сети 96.908%\n"
     ]
    }
   ],
   "source": [
    "plt.figure(figsize = (16,5))\n",
    "for i, err in enumerate(batch_errors):\n",
    "    plt.plot(err, color=colors[i])\n",
    "    \n",
    "\n",
    "plt.xlabel('Обучение')\n",
    "plt.ylabel('Ошибка')\n",
    "plt.show()\n",
    "\n",
    "print(f\"Аккуратность нейронной сети {(1 - errors[-1]) * 100:.3f}%\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "f9a5be1e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Аккуратность нейронной сети на тесте: 98.667%\n"
     ]
    }
   ],
   "source": [
    "y_pred = nn_model.predict(X_test)\n",
    "acc = (1 - np.mean(np.abs((y_test - y_pred)))) * 100\n",
    "print(f\"Аккуратность нейронной сети на тесте: {acc:.3f}%\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.6"
  },
  "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": 5
}