kay-gan.ipynb

{
 "cells": [
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "import logging\nlogger = logging.getLogger()\nlogger.setLevel(logging.DEBUG)",
   "execution_count": 1,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "data_path = \"/media/Data1/LukasG/KayGan/train_data.jblib\"\nfake_data_path = \"/media/Data1/LukasG/KayGan/train_data_fake.jblib\"\nfs = 250\nchannels = ['Fp1','Fp2','F7','F3','Fz','F4','F8','T7','C3','Cz','C4','T8','P7',\n            'P3','Pz','P4','P8','O1','O2','M1','M2']\nleft_hand = 0\nright_hand = 1",
   "execution_count": 96,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "import sys",
   "execution_count": 3,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "sys.path.insert(1, \"/home/lukasg/code/braindecode/\")",
   "execution_count": 4,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "from brainfeatures.data_set.abstract_data_set import DataSet",
   "execution_count": 5,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "import pandas as pd",
   "execution_count": 6,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "import joblib",
   "execution_count": 7,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "class KayGan(DataSet):\n    def __init__(self, path, subset, channel_names):\n        with open(path, 'rb') as fo:  \n            self.loaded = joblib.load(fo)[subset]\n        self.fs = 250\n        self.subset = subset\n            \n    def __getitem__(self, idx):\n        x = self.loaded.X[idx]\n        return pd.DataFrame(x, index=channels), self.fs, self.loaded.y[idx]\n    \n    def __len__(self):\n        return len(self.loaded.X)",
   "execution_count": 8,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "def load_train_valid_test_set(data_path, channel_names):\n    train_set = KayGan(\n        path=data_path,\n        subset=\"train\",\n        channel_names=channels\n    )\n    valid_set = KayGan(\n        path=data_path,\n        subset=\"valid\",\n        channel_names=channels\n    )\n    test_set = KayGan(\n        path=data_path,\n        subset=\"test\",\n        channel_names=channels\n    )\n    return train_set, valid_set, test_set",
   "execution_count": 71,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "train_set, valid_set, test_set = load_train_valid_test_set(data_path, channels)",
   "execution_count": 72,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "len(train_set), len(valid_set), len(test_set)",
   "execution_count": 73,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 73,
     "data": {
      "text/plain": "(358, 90, 80)"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "fake_train_set, fake_valid_set, fake_test_set = load_train_valid_test_set(fake_data_path, channels)",
   "execution_count": 98,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "len(fake_train_set), len(fake_valid_set), len(fake_test_set)",
   "execution_count": 99,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 99,
     "data": {
      "text/plain": "(358, 90, 80)"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "x, fs, y = train_set[0]",
   "execution_count": 14,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "x.shape, fs, y",
   "execution_count": 15,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 15,
     "data": {
      "text/plain": "((21, 384), 250, 1)"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "x.shape[-1]/fs",
   "execution_count": 16,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 16,
     "data": {
      "text/plain": "1.536"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "from brainfeatures.feature_generation.generate_features import default_feature_generation_params",
   "execution_count": 74,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "default_feature_generation_params.update({\"epoch_duration_s\": x.shape[-1]/fs})",
   "execution_count": 75,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "from brainfeatures.experiment.experiment import Experiment",
   "execution_count": 26,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "def generate_features(list_of_data_sets):\n    feats = []\n    for ds in list_of_data_sets:\n        exp = Experiment(\n            devel_set=ds,\n            estimator=None,\n            preproc_function=None,\n            feature_generation_params=default_feature_generation_params,\n            n_jobs=50\n        )\n        exp.run()\n        feats.append(pd.concat(exp._features[\"devel\"], ignore_index=True))\n    return feats",
   "execution_count": 76,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "train_feats, valid_feats, test_feats = generate_features([train_set, valid_set, test_set])",
   "execution_count": 77,
   "outputs": [
    {
     "output_type": "stream",
     "text": "INFO:root:Started on 2019-03-07 at 16:32:13.629218\nINFO:root:Loading devel (preprocessed)\nINFO:root:Generating features (devel)\nINFO:root:Finished on 2019-03-07 at 16:32:36.524810.\nINFO:root:Started on 2019-03-07 at 16:32:37.095011\nINFO:root:Loading devel (preprocessed)\nINFO:root:Generating features (devel)\nINFO:root:Finished on 2019-03-07 at 16:32:41.971529.\nINFO:root:Started on 2019-03-07 at 16:32:42.064338\nINFO:root:Loading devel (preprocessed)\nINFO:root:Generating features (devel)\nINFO:root:Finished on 2019-03-07 at 16:32:47.496739.\n",
     "name": "stderr"
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "train_feats.shape, valid_feats.shape, test_feats.shape",
   "execution_count": 78,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 78,
     "data": {
      "text/plain": "((358, 8463), (90, 8463), (80, 8463))"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "train_feats[:3]",
   "execution_count": 181,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 181,
     "data": {
      "text/plain": "   cwt_bounded_variation_a7_Fp1  cwt_bounded_variation_a7_Fp2  \\\n0                     12.491957                     14.566269   \n1                     49.169237                     18.202797   \n2                      9.294037                      7.581290   \n\n   cwt_bounded_variation_a7_F7  cwt_bounded_variation_a7_F3  \\\n0                    23.488305                    13.358394   \n1                    16.328832                    36.492521   \n2                    33.104700                    18.012985   \n\n   cwt_bounded_variation_a7_Fz  cwt_bounded_variation_a7_F4  \\\n0                    10.170716                    20.240715   \n1                    13.102106                    14.456780   \n2                     9.382596                    10.728926   \n\n   cwt_bounded_variation_a7_F8  cwt_bounded_variation_a7_T7  \\\n0                    48.457450                    72.134396   \n1                    50.510341                    56.999656   \n2                    13.532158                    84.702942   \n\n   cwt_bounded_variation_a7_C3  cwt_bounded_variation_a7_Cz  \\\n0                    17.169753                    12.602528   \n1                    17.496603                    28.061434   \n2                     8.270255                     8.210823   \n\n                 ...                 time_zero-crossing-derivative_T8  \\\n0                ...                                             95.0   \n1                ...                                             92.0   \n2                ...                                             94.0   \n\n   time_zero-crossing-derivative_P7  time_zero-crossing-derivative_P3  \\\n0                              87.0                             100.0   \n1                              93.0                              97.0   \n2                              94.0                              96.0   \n\n   time_zero-crossing-derivative_Pz  time_zero-crossing-derivative_P4  \\\n0                             102.0                              95.0   \n1                              92.0                              99.0   \n2                             100.0                              91.0   \n\n   time_zero-crossing-derivative_P8  time_zero-crossing-derivative_O1  \\\n0                              93.0                             101.0   \n1                              89.0                             100.0   \n2                              90.0                              95.0   \n\n   time_zero-crossing-derivative_O2  time_zero-crossing-derivative_M1  \\\n0                              98.0                              83.0   \n1                              98.0                              87.0   \n2                              94.0                             102.0   \n\n   time_zero-crossing-derivative_M2  \n0                              94.0  \n1                              88.0  \n2                              92.0  \n\n[3 rows x 8463 columns]",
      "text/html": "<div>\n<style scoped>\n    .dataframe tbody tr th:only-of-type {\n        vertical-align: middle;\n    }\n\n    .dataframe tbody tr th {\n        vertical-align: top;\n    }\n\n    .dataframe thead th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>cwt_bounded_variation_a7_Fp1</th>\n      <th>cwt_bounded_variation_a7_Fp2</th>\n      <th>cwt_bounded_variation_a7_F7</th>\n      <th>cwt_bounded_variation_a7_F3</th>\n      <th>cwt_bounded_variation_a7_Fz</th>\n      <th>cwt_bounded_variation_a7_F4</th>\n      <th>cwt_bounded_variation_a7_F8</th>\n      <th>cwt_bounded_variation_a7_T7</th>\n      <th>cwt_bounded_variation_a7_C3</th>\n      <th>cwt_bounded_variation_a7_Cz</th>\n      <th>...</th>\n      <th>time_zero-crossing-derivative_T8</th>\n      <th>time_zero-crossing-derivative_P7</th>\n      <th>time_zero-crossing-derivative_P3</th>\n      <th>time_zero-crossing-derivative_Pz</th>\n      <th>time_zero-crossing-derivative_P4</th>\n      <th>time_zero-crossing-derivative_P8</th>\n      <th>time_zero-crossing-derivative_O1</th>\n      <th>time_zero-crossing-derivative_O2</th>\n      <th>time_zero-crossing-derivative_M1</th>\n      <th>time_zero-crossing-derivative_M2</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>12.491957</td>\n      <td>14.566269</td>\n      <td>23.488305</td>\n      <td>13.358394</td>\n      <td>10.170716</td>\n      <td>20.240715</td>\n      <td>48.457450</td>\n      <td>72.134396</td>\n      <td>17.169753</td>\n      <td>12.602528</td>\n      <td>...</td>\n      <td>95.0</td>\n      <td>87.0</td>\n      <td>100.0</td>\n      <td>102.0</td>\n      <td>95.0</td>\n      <td>93.0</td>\n      <td>101.0</td>\n      <td>98.0</td>\n      <td>83.0</td>\n      <td>94.0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>49.169237</td>\n      <td>18.202797</td>\n      <td>16.328832</td>\n      <td>36.492521</td>\n      <td>13.102106</td>\n      <td>14.456780</td>\n      <td>50.510341</td>\n      <td>56.999656</td>\n      <td>17.496603</td>\n      <td>28.061434</td>\n      <td>...</td>\n      <td>92.0</td>\n      <td>93.0</td>\n      <td>97.0</td>\n      <td>92.0</td>\n      <td>99.0</td>\n      <td>89.0</td>\n      <td>100.0</td>\n      <td>98.0</td>\n      <td>87.0</td>\n      <td>88.0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>9.294037</td>\n      <td>7.581290</td>\n      <td>33.104700</td>\n      <td>18.012985</td>\n      <td>9.382596</td>\n      <td>10.728926</td>\n      <td>13.532158</td>\n      <td>84.702942</td>\n      <td>8.270255</td>\n      <td>8.210823</td>\n      <td>...</td>\n      <td>94.0</td>\n      <td>94.0</td>\n      <td>96.0</td>\n      <td>100.0</td>\n      <td>91.0</td>\n      <td>90.0</td>\n      <td>95.0</td>\n      <td>94.0</td>\n      <td>102.0</td>\n      <td>92.0</td>\n    </tr>\n  </tbody>\n</table>\n<p>3 rows × 8463 columns</p>\n</div>"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "fake_train_feats, fake_valid_feats, fake_test_feats = generate_features([fake_train_set, fake_valid_set, \n                                                                         fake_test_set])",
   "execution_count": 100,
   "outputs": [
    {
     "output_type": "stream",
     "text": "INFO:root:Started on 2019-03-07 at 16:37:41.840862\nINFO:root:Loading devel (preprocessed)\nINFO:root:Generating features (devel)\nINFO:root:Finished on 2019-03-07 at 16:37:59.955030.\nINFO:root:Started on 2019-03-07 at 16:38:00.470538\nINFO:root:Loading devel (preprocessed)\nINFO:root:Generating features (devel)\nINFO:root:Finished on 2019-03-07 at 16:38:05.723335.\nINFO:root:Started on 2019-03-07 at 16:38:05.823948\nINFO:root:Loading devel (preprocessed)\nINFO:root:Generating features (devel)\nINFO:root:Finished on 2019-03-07 at 16:38:11.449096.\n",
     "name": "stderr"
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "fake_train_feats.shape, fake_valid_feats.shape, fake_test_feats.shape",
   "execution_count": 102,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 102,
     "data": {
      "text/plain": "((358, 8463), (90, 8463), (80, 8463))"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "from sklearn.ensemble import RandomForestClassifier",
   "execution_count": 173,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "from sklearn.metrics import accuracy_score",
   "execution_count": 174,
   "outputs": []
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "def evaluate(clf, train_feats, train_y, valid_feats, valid_y, test_feats, test_y):\n    clf.fit(train_feats, train_y)\n    \n    train_preds = clf.predict(train_feats)\n    train_acc = accuracy_score(train_y, train_preds)\n    \n    valid_preds = clf.predict(valid_feats)\n    valid_acc = accuracy_score(valid_y, valid_preds)\n    \n    test_preds = clf.predict(test_feats)\n    test_acc = accuracy_score(test_y, test_preds)\n    return {\"train\": train_acc, \"valid\": valid_acc, \"test\": test_acc}",
   "execution_count": 175,
   "outputs": []
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "real data"
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "evaluate(RandomForestClassifier(n_estimators=100, random_state=0),\n         train_feats, [y for x ,fs, y in train_set], \n         valid_feats, [y for x, fs, y in valid_set], \n         test_feats, [y for x, fs, y in test_set])",
   "execution_count": 176,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 176,
     "data": {
      "text/plain": "{'train': 1.0, 'valid': 0.6111111111111112, 'test': 0.625}"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "fake data"
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "evaluate(RandomForestClassifier(n_estimators=100, random_state=0),\n         fake_train_feats, [y for x ,fs, y in fake_train_set], \n         fake_valid_feats, [y for x, fs, y in fake_valid_set], \n         fake_test_feats, [y for x, fs, y in fake_test_set])",
   "execution_count": 177,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 177,
     "data": {
      "text/plain": "{'train': 1.0, 'valid': 0.7, 'test': 0.675}"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "train on real data, predict fake data"
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "evaluate(RandomForestClassifier(n_estimators=100, random_state=0),\n         train_feats, [y for x ,fs, y in train_set], \n         fake_valid_feats, [y for x, fs, y in fake_valid_set], \n         fake_test_feats, [y for x, fs, y in fake_test_set])",
   "execution_count": 178,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 178,
     "data": {
      "text/plain": "{'train': 1.0, 'valid': 0.6111111111111112, 'test': 0.5875}"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {},
   "cell_type": "markdown",
   "source": "train on fake data, predict real data"
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "evaluate(RandomForestClassifier(n_estimators=100, random_state=0),\n         fake_train_feats, [y for x ,fs, y in fake_train_set], \n         valid_feats, [y for x, fs, y in valid_set], \n         test_feats, [y for x, fs, y in test_set])",
   "execution_count": 179,
   "outputs": [
    {
     "output_type": "execute_result",
     "execution_count": 179,
     "data": {
      "text/plain": "{'train': 1.0, 'valid': 0.6111111111111112, 'test': 0.5375}"
     },
     "metadata": {}
    }
   ]
  },
  {
   "metadata": {
    "trusted": true
   },
   "cell_type": "code",
   "source": "",
   "execution_count": null,
   "outputs": []
  }
 ],
 "metadata": {
  "language_info": {
   "pygments_lexer": "ipython3",
   "codemirror_mode": {
    "version": 3,
    "name": "ipython"
   },
   "version": "3.6.7",
   "nbconvert_exporter": "python",
   "mimetype": "text/x-python",
   "name": "python",
   "file_extension": ".py"
  },
  "kernelspec": {
   "name": "brainfeatures",
   "display_name": "brainfeatures",
   "language": "python"
  },
  "gist_id": "e5f7615fafd65b4ab54c77be2c7df52c"
 },
 "nbformat": 4,
 "nbformat_minor": 2
}