presentation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "presentation.ipynb",
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ra2jatEot3k2",
        "colab_type": "text"
      },
      "source": [
        "# **Introduction**\n",
        "\n",
        "\n",
        "My name is Shiva Shamloo. I am a master student in the Concordia University. I joined the BrainHack school in order to gain some experience with neuroimage data and to find out more about the algorithms and the methods used in the neuroscience. For my project I decided to work on brain decoding in order to get familiarize with the fMRI data."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "86yHggov038v",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 224
        },
        "outputId": "8e295d36-e42e-44eb-be46-7ce029247288"
      },
      "source": [
        "!pip install nilearn"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting nilearn\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/b9/c2/f5f1bdd37a3da28b3b34305e4ba27cce468db6073998d62a38abd0e281da/nilearn-0.6.2-py3-none-any.whl (2.5MB)\n",
            "\u001b[K     |████████████████████████████████| 2.5MB 2.7MB/s \n",
            "\u001b[?25hRequirement already satisfied: sklearn in /usr/local/lib/python3.6/dist-packages (from nilearn) (0.0)\n",
            "Requirement already satisfied: scikit-learn>=0.19 in /usr/local/lib/python3.6/dist-packages (from nilearn) (0.22.2.post1)\n",
            "Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.6/dist-packages (from nilearn) (0.15.1)\n",
            "Requirement already satisfied: nibabel>=2.0.2 in /usr/local/lib/python3.6/dist-packages (from nilearn) (3.0.2)\n",
            "Requirement already satisfied: numpy>=1.11 in /usr/local/lib/python3.6/dist-packages (from nilearn) (1.18.4)\n",
            "Requirement already satisfied: scipy>=0.19 in /usr/local/lib/python3.6/dist-packages (from nilearn) (1.4.1)\n",
            "Installing collected packages: nilearn\n",
            "Successfully installed nilearn-0.6.2\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "almvsi790k1_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.multiclass import OneVsOneClassifier, OneVsRestClassifier\n",
        "from sklearn.gaussian_process import GaussianProcessClassifier\n",
        "from sklearn.ensemble import GradientBoostingClassifier\n",
        "from sklearn.gaussian_process.kernels import RBF\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.model_selection import cross_val_score\n",
        "from sklearn.neighbors import KNeighborsClassifier\n",
        "from sklearn.linear_model import LogisticRegression\n",
        "from nilearn.input_data import NiftiMasker\n",
        "from sklearn.metrics import classification_report\n",
        "from matplotlib import pyplot as plt\n",
        "from nilearn.plotting import plot_stat_map, show\n",
        "from nilearn import datasets\n",
        "from sklearn.svm import SVC\n",
        "from nilearn import image\n",
        "import pandas as pd"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7kPx1oYp1MbS",
        "colab_type": "text"
      },
      "source": [
        "# Loading the data\n",
        "\n",
        "For this rpoject I will be using the Haxby dataset which is already preprocessed and it is available in the nilearn. The dataset is consist of the fMRI data of 6 people: 5 women and one man. For this project I will be using ventral temporal cortex for training the model. \n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Xh5kGNnR2OEa",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "haxby_dataset = datasets.fetch_haxby()\n",
        "\n",
        "fmri_filename = haxby_dataset.func[0]\n",
        "\n",
        "mask_filename = haxby_dataset.mask_vt[0]\n",
        "\n",
        "masker = NiftiMasker(mask_img=mask_filename, standardize=True)\n",
        "\n",
        "behavioral = pd.read_csv(haxby_dataset.session_target[0], sep=\" \")\n",
        "\n",
        "conditions = behavioral['labels']\n",
        "print(conditions.unique())\n",
        "\n",
        "fmri_masked = masker.fit_transform(fmri_filename)\n",
        "\n",
        "fmri_train, fmri_test, conditions_train, conditions_test = train_test_split(fmri_masked, conditions, test_size=0.2, random_state=0)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vvGrs3Ug_bkK",
        "colab_type": "text"
      },
      "source": [
        "# Building the Decoders\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aV1RcDczOsMC",
        "colab_type": "text"
      },
      "source": [
        "# Logistic Regression\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "q9v-1wrYO1jo",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 911
        },
        "outputId": "8c461782-debd-4fea-f298-4e14c1f0bd14"
      },
      "source": [
        "Log_reg_ovr = LogisticRegression(random_state=0, max_iter=500, multi_class='ovr').fit(fmri_train, conditions_train)\n",
        "Log_reg_ovo = LogisticRegression(random_state=0, max_iter=500, multi_class='multinomial').fit(fmri_train, conditions_train)\n",
        "\n",
        "log_reg_ovr_prediction=Log_reg_ovr.predict(fmri_test)\n",
        "log_reg_ovo_prediction=Log_reg_ovo.predict(fmri_test)\n",
        "\n",
        "Log_reg_ovr_report=classification_report(conditions_test,log_reg_ovr_prediction)\n",
        "Log_reg_ovo_report=classification_report(conditions_test,log_reg_ovo_prediction)\n",
        "\n",
        "print('One vs. One: '+str(Log_reg_ovo_report))\n",
        "print('One vs. Rest: '+str(Log_reg_ovr_report))\n",
        "\n",
        "cv_scores_ovo = cross_val_score(Log_reg_ovr, fmri_train, conditions_train, cv=5, verbose=1)\n",
        "cv_scores_ova = cross_val_score(Log_reg_ovo, fmri_train, conditions_train, cv=5, verbose=1)\n",
        "\n",
        "print('Cross Validation OvO:', cv_scores_ovo.mean())\n",
        "print('Cross Validation OvR:', cv_scores_ova.mean())\n",
        "\n",
        "plt.figure(figsize=(6, 4))\n",
        "plt.plot(cv_scores_ovo, label='Cross validation OvO')\n",
        "plt.plot(cv_scores_ova, label='Cross validation OvR')\n",
        "plt.legend(loc='best', frameon=False)\n",
        "show()\n"
      ],
      "execution_count": 72,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "One vs. One:               precision    recall  f1-score   support\n",
            "\n",
            "      bottle       0.83      0.91      0.87        22\n",
            "         cat       0.85      0.94      0.89        18\n",
            "       chair       0.79      0.95      0.86        20\n",
            "        face       1.00      0.88      0.94        25\n",
            "       house       1.00      0.89      0.94        19\n",
            "        rest       0.95      0.96      0.95       117\n",
            "    scissors       0.95      0.82      0.88        22\n",
            "scrambledpix       0.96      1.00      0.98        25\n",
            "        shoe       0.95      0.87      0.91        23\n",
            "\n",
            "    accuracy                           0.93       291\n",
            "   macro avg       0.92      0.91      0.91       291\n",
            "weighted avg       0.93      0.93      0.93       291\n",
            "\n",
            "One vs. Rest:               precision    recall  f1-score   support\n",
            "\n",
            "      bottle       0.80      0.73      0.76        22\n",
            "         cat       0.81      0.94      0.87        18\n",
            "       chair       0.79      0.75      0.77        20\n",
            "        face       1.00      0.84      0.91        25\n",
            "       house       0.85      0.89      0.87        19\n",
            "        rest       0.94      0.97      0.95       117\n",
            "    scissors       0.86      0.86      0.86        22\n",
            "scrambledpix       1.00      0.96      0.98        25\n",
            "        shoe       0.83      0.87      0.85        23\n",
            "\n",
            "    accuracy                           0.90       291\n",
            "   macro avg       0.88      0.87      0.87       291\n",
            "weighted avg       0.90      0.90      0.90       291\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n",
            "[Parallel(n_jobs=1)]: Done   5 out of   5 | elapsed:    4.4s finished\n",
            "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Cross Validation OvO: 0.898357259138671\n",
            "Cross Validation OvR: 0.9026713038330619\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=1)]: Done   5 out of   5 | elapsed:    4.2s finished\n"
          ],
          "name": "stderr"
        },
        {
          "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": "2AhY-zzfO7wL",
        "colab_type": "text"
      },
      "source": [
        "# Support Vector Machines"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sJIHCy30_vzg",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 911
        },
        "outputId": "473f0766-4a18-4afa-954a-4a6c8dd4e5f9"
      },
      "source": [
        "svc = SVC(kernel='linear')\n",
        "svo=OneVsOneClassifier(svc)\n",
        "sva=OneVsRestClassifier(svc)\n",
        "\n",
        "svo.fit(fmri_train,conditions_train)\n",
        "sva.fit(fmri_train,conditions_train)\n",
        "\n",
        "\n",
        "svo_prediction = svo.predict(fmri_test)\n",
        "svo_report=classification_report(conditions_test,svo_prediction)\n",
        "\n",
        "sva_prediction = sva.predict(fmri_test)\n",
        "sva_report=classification_report(conditions_test,sva_prediction)\n",
        "\n",
        "print('One vs. One: ' + str(svo_report))\n",
        "print('One vs. Rest: '+ str(sva_report))\n",
        "\n",
        "cv_scores_ovo = cross_val_score(svo, fmri_train, conditions_train, cv=5, verbose=1)\n",
        "cv_scores_ova = cross_val_score(sva, fmri_train, conditions_train, cv=5, verbose=1)\n",
        "\n",
        "print('Cross Validation OvO:', cv_scores_ovo.mean())\n",
        "print('Cross Validation OvR:', cv_scores_ova.mean())\n",
        "\n",
        "plt.figure(figsize=(6, 4))\n",
        "plt.plot(cv_scores_ovo, label='Cross validation OvO')\n",
        "plt.plot(cv_scores_ova, label='Cross validation OvR')\n",
        "plt.legend(loc='best', frameon=False)\n",
        "show()\n"
      ],
      "execution_count": 73,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "One vs. One:               precision    recall  f1-score   support\n",
            "\n",
            "      bottle       0.83      0.91      0.87        22\n",
            "         cat       0.80      0.89      0.84        18\n",
            "       chair       0.81      0.85      0.83        20\n",
            "        face       1.00      0.84      0.91        25\n",
            "       house       0.94      0.89      0.92        19\n",
            "        rest       0.96      0.96      0.96       117\n",
            "    scissors       0.91      0.91      0.91        22\n",
            "scrambledpix       0.88      0.92      0.90        25\n",
            "        shoe       0.95      0.91      0.93        23\n",
            "\n",
            "    accuracy                           0.92       291\n",
            "   macro avg       0.90      0.90      0.90       291\n",
            "weighted avg       0.92      0.92      0.92       291\n",
            "\n",
            "One vs. Rest:               precision    recall  f1-score   support\n",
            "\n",
            "      bottle       0.77      0.77      0.77        22\n",
            "         cat       0.73      0.89      0.80        18\n",
            "       chair       0.78      0.70      0.74        20\n",
            "        face       1.00      0.84      0.91        25\n",
            "       house       0.85      0.89      0.87        19\n",
            "        rest       0.94      0.96      0.95       117\n",
            "    scissors       0.78      0.82      0.80        22\n",
            "scrambledpix       0.92      0.96      0.94        25\n",
            "        shoe       0.85      0.74      0.79        23\n",
            "\n",
            "    accuracy                           0.88       291\n",
            "   macro avg       0.85      0.84      0.84       291\n",
            "weighted avg       0.88      0.88      0.88       291\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n",
            "[Parallel(n_jobs=1)]: Done   5 out of   5 | elapsed:    6.9s finished\n",
            "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Cross Validation OvO: 0.8879902323516353\n",
            "Cross Validation OvR: 0.8811269794287406\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=1)]: Done   5 out of   5 | elapsed:    6.8s finished\n"
          ],
          "name": "stderr"
        },
        {
          "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": "qhwCL6qkPGbo",
        "colab_type": "text"
      },
      "source": [
        "# Gaussian Process Classifier"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "W_AYj4EFLSlH",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 666
        },
        "outputId": "d33fef9a-4f20-4419-c4fb-54e558b673b6"
      },
      "source": [
        "gpc_o = GaussianProcessClassifier(multi_class = 'one_vs_one').fit(fmri_train, conditions_train)\n",
        "gpc_r = GaussianProcessClassifier(multi_class = 'one_vs_rest').fit(fmri_train, conditions_train)\n",
        "\n",
        "gpc_o_prediction= gpc_o.predict(fmri_test)\n",
        "gpc_r_prediction= gpc_r.predict(fmri_test)\n",
        "\n",
        "gpc_o_report=classification_report(conditions_test,gpc_o_prediction)\n",
        "gpc_r_report=classification_report(conditions_test,gpc_r_prediction)\n",
        "\n",
        "print('One vs. One: ' + str(gpc_o_report))\n",
        "print('One vs. Rest: ' + str(gpc_r_report))\n",
        "\n",
        "cv_scores_ovo = cross_val_score(gpc_o, fmri_train, conditions_train, cv=5, verbose=1)\n",
        "cv_scores_ova = cross_val_score(gpc_r, fmri_train, conditions_train, cv=5, verbose=1)\n",
        "\n",
        "print('Cross Validation OvO:', cv_scores_ovo.mean())\n",
        "print('Cross Validation OvR:', cv_scores_ova.mean())\n",
        "\n",
        "\n",
        "plt.figure(figsize=(6, 4))\n",
        "plt.plot(cv_scores_ovo, label='Cross validation OvO')\n",
        "plt.plot(cv_scores_ova, label='Cross validation OvR')\n",
        "plt.legend(loc='best', frameon=False)\n",
        "show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n",
            "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "One vs. One:               precision    recall  f1-score   support\n",
            "\n",
            "      bottle       0.83      0.86      0.84        22\n",
            "         cat       0.62      1.00      0.77        18\n",
            "       chair       0.77      0.85      0.81        20\n",
            "        face       0.86      0.72      0.78        25\n",
            "       house       0.90      0.95      0.92        19\n",
            "        rest       0.97      0.84      0.90       117\n",
            "    scissors       0.90      0.86      0.88        22\n",
            "scrambledpix       0.80      0.96      0.87        25\n",
            "        shoe       0.75      0.78      0.77        23\n",
            "\n",
            "    accuracy                           0.86       291\n",
            "   macro avg       0.82      0.87      0.84       291\n",
            "weighted avg       0.87      0.86      0.86       291\n",
            "\n",
            "One vs. Rest:               precision    recall  f1-score   support\n",
            "\n",
            "      bottle       0.00      0.00      0.00        22\n",
            "         cat       0.00      0.00      0.00        18\n",
            "       chair       0.00      0.00      0.00        20\n",
            "        face       0.00      0.00      0.00        25\n",
            "       house       0.00      0.00      0.00        19\n",
            "        rest       0.00      0.00      0.00       117\n",
            "    scissors       0.00      0.00      0.00        22\n",
            "scrambledpix       0.00      0.00      0.00        25\n",
            "        shoe       0.08      1.00      0.15        23\n",
            "\n",
            "    accuracy                           0.08       291\n",
            "   macro avg       0.01      0.11      0.02       291\n",
            "weighted avg       0.01      0.08      0.01       291\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "[Parallel(n_jobs=1)]: Done   5 out of   5 | elapsed:   35.4s finished\n",
            "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "awmk6S5sPiT5",
        "colab_type": "text"
      },
      "source": [
        "# Gradient Boosting Classifier\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1ZUDEKq8Pnmy",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 289
        },
        "outputId": "dc6d2500-6ace-43ed-96e3-c54bac1fc553"
      },
      "source": [
        "gbc = GradientBoostingClassifier(random_state=0)\n",
        "gbc.fit(fmri_train, conditions_train)\n",
        "gbc_prediction=gbc.predict(fmri_test)\n",
        "gbc_report=classification_report(conditions_test,gbc_prediction)\n",
        "\n",
        "print(gbc_report)\n",
        "\n",
        "cv_scores = cross_val_score(gbc, fmri_train, conditions_train, cv=5, verbose=1)\n",
        "\n",
        "\n",
        "print('Cross Validation:', cv_scores.mean())\n",
        "\n",
        "\n",
        "plt.figure(figsize=(6, 4))\n",
        "plt.plot(cv_scores, label='Cross validation')\n",
        "plt.legend(loc='best', frameon=False)\n",
        "show()\n"
      ],
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "      bottle       0.62      0.59      0.60        22\n",
            "         cat       0.55      0.61      0.58        18\n",
            "       chair       0.48      0.60      0.53        20\n",
            "        face       0.78      0.56      0.65        25\n",
            "       house       1.00      0.79      0.88        19\n",
            "        rest       0.86      0.96      0.91       117\n",
            "    scissors       0.56      0.41      0.47        22\n",
            "scrambledpix       0.75      0.60      0.67        25\n",
            "        shoe       0.54      0.61      0.57        23\n",
            "\n",
            "    accuracy                           0.74       291\n",
            "   macro avg       0.68      0.64      0.65       291\n",
            "weighted avg       0.74      0.74      0.73       291\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ce1F6t01Am3g",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}