LogisticRegression.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "LogisticRegression.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyOJGZznjUKm8A5w4ndE2Uwf",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/AntonThai2022/ML/blob/main/LogisticRegression.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import math\n",
        "import tensorflow as tf\n",
        "from matplotlib import pyplot as plt"
      ],
      "metadata": {
        "id": "3OOO3vFjs2Va"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Создаём по 100 точек для бинарной классификации (кол-во точек можно менять)"
      ],
      "metadata": {
        "id": "qSyDu0Imr2P-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "N = 200\n",
        "x_zeros = np.random.multivariate_normal(mean=np.array((-1,-1)), cov=.1*np.eye(2), size=(int(N/2),))\n",
        "y_zeros = np.zeros((int(N/2),))\n",
        "x_ones = np.random.multivariate_normal(mean=np.array((1,1)), cov=.1*np.eye(2), size=(int(N/2),))\n",
        "y_ones = np.ones((int(N/2), ))\n",
        "x_np = np.vstack([x_zeros, x_ones])\n",
        "y_np = np.concatenate([y_zeros, y_ones])"
      ],
      "metadata": {
        "id": "X1P_YPJzMg5N"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Визуализируем точки"
      ],
      "metadata": {
        "id": "-2j_P7r2x8mX"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        },
        "id": "KZKKgYQ0rPhU",
        "outputId": "9825f709-8819-4c1d-9914-bcd2e4b135c2"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ],
      "source": [
        "#Визуализируем по оси все первые координаты каждой точки, а по другой оси все вторые координаты каждой точки \n",
        "plt.scatter(x_np[:,0], x_np[:,1])\n",
        "plt.title(\"Sample visualization\")\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Определяем модель логистической регрессии"
      ],
      "metadata": {
        "id": "w3MJ5JK2IKMw"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "tf.compat.v1.disable_eager_execution() \n",
        "with tf.name_scope(\"placaholders\"):\n",
        "  x = tf.compat.v1.placeholder(tf.float32, (N,2))\n",
        "  y = tf.compat.v1.placeholder(tf.float32, (N,))\n",
        "with tf.name_scope(\"weights\"):\n",
        "  W = tf.compat.v1.Variable(tf.compat.v1.random_normal((2,1)))\n",
        "  b = tf.compat.v1.Variable(tf.compat.v1.random_normal((1,)))\n",
        "with tf.name_scope(\"prediction\"):\n",
        "  y_logit = tf.compat.v1.squeeze(tf.compat.v1.matmul(x,W) + b)\n",
        "  y_one_prob = tf.compat.v1.sigmoid(y_logit)\n",
        "  y_pred = tf.round(y_one_prob)\n",
        "\n",
        "with tf.name_scope(\"loss\"):\n",
        "  entropy = tf.nn.sigmoid_cross_entropy_with_logits(logits=y_logit, labels=y)\n",
        "  l = tf.reduce_sum(entropy)\n",
        "with tf.name_scope(\"optim\"):\n",
        "  train_op = tf.compat.v1.train.AdamOptimizer(.01).minimize(l)"
      ],
      "metadata": {
        "id": "D-Rhi0wHH_J7"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Тренировка ллогистической регрессии"
      ],
      "metadata": {
        "id": "Yu68TynRQAdp"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "n_steps = 1000\n",
        "with tf.compat.v1.Session() as sess:\n",
        "  sess.run(tf.compat.v1.global_variables_initializer())\n",
        "  for i in range(n_steps):\n",
        "    summary, loss = sess.run([train_op, l], feed_dict={x: x_np, y: y_np})\n",
        "    if(i % 100 == 0):\n",
        "      print(\"потеря: %f\" % loss) \n",
        "  W_value = W.eval() \n",
        "  b_value = b.eval()  "
      ],
      "metadata": {
        "id": "pA6EsPE6QGMp",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "aaf4d3d0-a8ef-4ddc-ccb9-00ac14cf0392"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "потеря: 126.711388\n",
            "потеря: 28.816486\n",
            "потеря: 12.277152\n",
            "потеря: 7.247461\n",
            "потеря: 4.986942\n",
            "потеря: 3.738020\n",
            "потеря: 2.955667\n",
            "потеря: 2.422764\n",
            "потеря: 2.037611\n",
            "потеря: 1.746794\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Визуализируем разделяющую линию, "
      ],
      "metadata": {
        "id": "Ke2MTfsNVqaz"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "#Получаем значения весов в виде обычного списка\n",
        "W_val = np.reshape(W_value, (1,2))[0]\n",
        "#И смещение\n",
        "b_val = b_value[0]\n",
        "#Определяем сигмоиду\n",
        "def sigmoid(x):\n",
        "  return 1 / (1 + math.exp(-x))\n",
        "\n",
        "#Теперь перебирам кучу точек из интереующеей области, и те из них, на которых регрессия выдаёт результат\n",
        "#близкий к 0.5 будут являться толчками линии, которая и разделяет наши классы\n",
        "x1, x2 = [], []\n",
        "for x1_test in np.linspace(-1.5,1.5, 100):\n",
        "  for x2_test in np.linspace(-1.5,1.5, 100):\n",
        "    #попробуй ради интереса поменяй местами индексы у весов\n",
        "    z = sigmoid(x1_test*W_val[0] + x2_test*W_val[1] + b_val)\n",
        "    if abs(z - 0.5) < 0.01:\n",
        "      x1.append(x1_test)\n",
        "      x2.append(x2_test)"
      ],
      "metadata": {
        "id": "7AAlilBgNcVA"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#Теперь делаем из наших списков numpy массивы и отрисовываем линию\n",
        "plt.scatter(x_np[:,0], x_np[:,1]) #точки данных\n",
        "plt.scatter(x1, x2) #теоретические точки, где регрессия уже может ошибиться, то есть линия, разделяющая данные\n",
        "plt.title(\"Logistic regression\")\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 281
        },
        "id": "7CwGIq7PR0WB",
        "outputId": "9a4bcad0-d644-4f14-9d9d-16ad058a56d9"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}