Exercise7b_v1.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Untitled1.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyOV+KUGwCXi94x7wXimGSbu",
      "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/Patschwemm/Container-Localisation/blob/main/Exercise7b_v1.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "FgmNOHY9PUQn"
      },
      "outputs": [],
      "source": [
        "%%capture\n",
        "!pip install igraph\n",
        "!pip install grakel"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "import igraph\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.metrics import accuracy_score\n",
        "from sklearn.svm import SVC \n",
        "import grakel\n",
        "import grakel.graph as gg\n",
        "from grakel.datasets import fetch_dataset\n",
        "from grakel.utils import cross_validate_Kfold_SVM\n",
        "from grakel.kernels import WeisfeilerLehman, VertexHistogram\n",
        "import pickle\n",
        "from pathlib import Path"
      ],
      "metadata": {
        "id": "RvH-4pF8U0vF"
      },
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Loads the MUTAG dataset\n",
        "NCI1 = fetch_dataset(\"NCI1\", verbose=False)\n",
        "DHFR = fetch_dataset(\"DHFR\", verbose=False)\n",
        "\n",
        "G = []\n",
        "y = []\n",
        "G.append(NCI1.data)\n",
        "y.append(NCI1.target)\n",
        "G.append(DHFR.data)\n",
        "y.append(DHFR.target)\n",
        "\n",
        "# Generates a list of kernel matrices using the Weisfeiler-Lehman subtree kernel\n",
        "# Each kernel matrix is generated by setting the number of iterations of the\n",
        "# kernel to a different value (from 2 to 7)\n",
        "# done for both datasets\n",
        "for G, y in zip(G,y):\n",
        "    Ks = list()\n",
        "    for i in range(1, 7):\n",
        "        gk = WeisfeilerLehman(n_iter=i, base_graph_kernel=VertexHistogram, normalize=True)\n",
        "        K = gk.fit_transform(G)\n",
        "        Ks.append(K)\n",
        "\n",
        "    # Performs 10-fold cross-validation over different kernels and the parameter C of\n",
        "    # SVM and repeats the experiment 10 times with different folds\n",
        "    accs = cross_validate_Kfold_SVM([Ks], y, n_iter=10)\n",
        "    print(\"Average accuracy:\", str(round(np.mean(accs[0])*100, 2)) + \"%\")\n",
        "    print(\"Standard deviation:\", str(round(np.std(accs[0])*100, 2)) + \"%\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8C6PhJH-Vy4s",
        "outputId": "c8ca9a22-c747-4d88-ff79-371b3e7dde5e"
      },
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Average accuracy: 84.7%\n",
            "Standard deviation: 0.22%\n",
            "Average accuracy: 81.03%\n",
            "Standard deviation: 1.18%\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Loads the MUTAG dataset\n",
        "NCI1 = fetch_dataset(\"NCI1\", verbose=False)\n",
        "DHFR = fetch_dataset(\"DHFR\", verbose=False)\n",
        "\n",
        "G = []\n",
        "y = []\n",
        "G.append(NCI1.data)\n",
        "y.append(NCI1.target)\n",
        "G.append(DHFR.data)\n",
        "y.append(DHFR.target)\n",
        "\n",
        "\n",
        "\n",
        "for G, y in zip(G,y):\n",
        "    # generate VertexHistogram Kernel\n",
        "    G_train, G_test, y_train, y_test = train_test_split(G, y, test_size=0.1, random_state=42)\n",
        "    gk = VertexHistogram( normalize=False)\n",
        "    K_train = gk.fit_transform(G_train)\n",
        "    K_test = gk.transform(G_test)\n",
        "\n",
        "    # Uses the SVM classifier to perform classification\n",
        "    clf = SVC(kernel=\"precomputed\")\n",
        "    clf.fit(K_train, y_train)\n",
        "    y_pred = clf.predict(K_test)\n",
        "\n",
        "    # Computes and prints the classification accuracy\n",
        "    acc = accuracy_score(y_test, y_pred)\n",
        "    print(\"Accuracy:\", str(round(acc*100, 2)) + \"%\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xxJs5468xto6",
        "outputId": "9b0c2e5f-5bea-4c17-f6b6-a9dab0eaed9d"
      },
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy: 65.94%\n",
            "Accuracy: 59.21%\n"
          ]
        }
      ]
    }
  ]
}