Jun_ML_Clusterization_HW_7_4.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Jun ML Clusterization HW 7.4.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "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": "3.6.3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/oObelix/DataScience/blob/master/Jun_ML_Clusterization_HW_7_4.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ToOBomb3PAB4"
      },
      "source": [
        "## Урок 4. Домашнее задание \n",
        "\n",
        "### 4.1 Домашнее задание начального уровня\n",
        "\n",
        "Потренируйтесь в программировании на python. Надо будет найти расстояние от каждой точки из массива $X$ до центра кластера, а потом просуммировать эти расстояния. Это будет реализацией библиотечной функции `sklearn.metrics.pairwise.euclidean_distances` из первого урока.  В итоге у Вас должно получиться число чуть более $16$ - это и будет расстояние.\n",
        "$$\n",
        "\\sum_{i=1}^{k} \\mid x_i - c \\mid\n",
        "$$\n",
        "\n",
        "где для одного обучающего примера x, состоящего из двух координат:\n",
        "$$\n",
        "\\mid x - c \\mid = (x_1-c_1)^2 + (x_2-c_2)^2\n",
        "$$\n",
        "\n",
        "**Примечание** Как вы уже знаете из урока про линейную регрессию, в этой формуле должен быть квадратный корень, но для скорости работы мы опускаем его.\n",
        "\n",
        "**Подсказка** Средствами библиотеки numpy реализуйте формулу нахождения суммы внутрикластерных расстояний"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nY5dFMW8PACP",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "3b6f40c0-e80f-4abe-ccb3-cc348b71d682"
      },
      "source": [
        "import numpy as np\n",
        "\n",
        "clust = np.array([\n",
        " [-0.5972191612445694, -0.5919098916910581],\n",
        " [-1.5838020751869848, 0.4743393635868491],\n",
        " [-1.892642118066139, -1.2770390481464395],\n",
        " [-1.021761443519372, -0.38446059106320013],\n",
        " [-0.628271339507516, -2.308149346281125],\n",
        " [-0.7180915776856387, 1.1805550909358404],\n",
        " [-1.543143767548152, -1.4163791359687334],\n",
        " [0.022103701018375554, -1.1279478858865397],\n",
        " [-0.7776518874305778, -0.4157532453316538],\n",
        " [-1.2073577296008344, -1.1308523658604184]\n",
        "])\n",
        "\n",
        "centroid = np.array([-0.9774245525274352, -0.032635425821084516])\n",
        "\n",
        "# -- ВАШ КОД ТУТ --\n",
        "\n",
        "\n",
        "# np.linalg.norm(clust - centroid)\n",
        "# np.linalg.norm(clust - centroid)**2\n",
        "\n",
        "sum([(xi[0]-centroid[0])**2 + (xi[1]-centroid[1])**2  for xi in clust])\n",
        "np.array([(xi[0]-centroid[0])**2 + (xi[1]-centroid[1])**2 for xi in clust]).sum()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "16.31211696438564"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "O2lFNZ_EPACV"
      },
      "source": [
        "### 4.2 Домашнее задание начального уровня  \n",
        "\n",
        "Обучите алгоритм k-средних для количества кластеров $n=3$ на том же датасете, который содержится в файле `data/clustering.pkl`. Визуализируйте результат:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GkTkaXo1PACW",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "outputId": "e2704339-8a2c-4df7-a9a8-cdb1d084ed83"
      },
      "source": [
        "# -- ВАШ КОД ТУТ --\n",
        "\n",
        "import pickle\n",
        "\n",
        "with open('sample_data/clustering.pkl', 'rb') as f:\n",
        "  clustering = pickle.load(f)\n",
        "\n",
        "clustering\n",
        "\n",
        "X = np.array(clustering['X'])\n",
        "# y = np.array(clustering['Y'])\n",
        "\n",
        "from sklearn.cluster import KMeans\n",
        "\n",
        "km_model = KMeans(n_clusters=3)\n",
        "km_model.fit(X)\n",
        "\n",
        "from matplotlib import pyplot as plt\n",
        "\n",
        "plt.scatter(X[:, 0], X[:, 1], c=km_model.labels_)\n",
        "plt.show()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kxSaSMsuPACd"
      },
      "source": [
        "### 4.3 Домашнее задание высокого уровня\n",
        "\n",
        "Мы узнали про эвристику, улучшающую алгоритм: запустить несколько раз алгоритм, запомнить, куда попали центроиды при каждом запуске, а потом усреднить координаты за все несколько запуска.\n",
        "\n",
        "Попробуйте применить эту эвристику на практике: \n",
        "\n",
        "* используйте 10 запусков алгоритма $k$-средних для $k=2$ с параметрами\n",
        "<pre>\n",
        "kmeans_model = KMeans(n_clusters=2, n_init=1, random_state=None, algorithm='full', max_iter=2)\n",
        "</pre>\n",
        "* в каждом запуске сохраняйте координаты центроидов общий массив $c$. См. `cluster_centers_ : array, [n_clusters, n_features]`\n",
        "* визуализируйте координаты точек из $c$ на фоне исходного множества точек\n",
        "\n",
        "Результат будет примерно такой:\n",
        "![several_init_knn](https://248006.selcdn.ru/public/DS.%20Block%202.%20M7/several_init_knn.png)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y7TYD_kRPACg",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "outputId": "27d43a98-362a-4075-ece6-1e3478e8589e"
      },
      "source": [
        "# -- ВАШ КОД ТУТ --\n",
        "\n",
        "plt.scatter(X[:, 0], X[:, 1], alpha=.4)\n",
        "\n",
        "cluster0_centr = []\n",
        "cluster1_centr = []\n",
        "for i in range(10):\n",
        "  km_model = KMeans(n_clusters=2, n_init=1, random_state=None, \n",
        "                    algorithm='full', max_iter=2)\n",
        "  km_model.fit(X)\n",
        "  # print(km_model.cluster_centers_)\n",
        "  cluster0_centr.append(km_model.cluster_centers_[0])\n",
        "  cluster1_centr.append(km_model.cluster_centers_[1])\n",
        "\n",
        "cluster0_centr = np.array(cluster0_centr)\n",
        "cluster1_centr = np.array(cluster1_centr)\n",
        "\n",
        "plt.scatter(cluster0_centr[:, 0], cluster0_centr[:, 1], c='red', s=80, marker='x')\n",
        "plt.scatter(cluster1_centr[:, 0], cluster1_centr[:, 1], c='orange', s=80, marker='x')\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rldfAxyWPACk"
      },
      "source": [
        "Мы выяснили, что один из недостатков алгоритма k-средних состоит в том, что количество кластеров надо задать до тренировки модели, но как же определить, сколько будет кластеров в данных? Для этой задачи существует эффективный алгоритм, о котором мы поговорим в следующем уроке."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZH1FEaT8T-pe"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}