visualisation.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "source": [
    "# Results visualization\n",
    "## imports"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "2ff99ab797ab804a"
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "initial_id",
   "metadata": {
    "collapsed": true,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:29.487953Z",
     "start_time": "2024-02-13T17:08:29.473729Z"
    }
   },
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "\n",
    "import numpy as np\n",
    "from scipy.stats import wilcoxon, false_discovery_control\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Load results and sanity check"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "e38e199408424350"
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "# Combination of columns that should lead to a unique result within each file\n",
    "unique_cols = ['pretrain_dataset', 'finetune_dataset', 'finetune_scheme', 'finetune_subject',\n",
    "               'finetune_fold', 'directory', 'ref_channel']"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:30.091841Z",
     "start_time": "2024-02-13T17:08:30.089222Z"
    }
   },
   "id": "7596d098876425f7",
   "execution_count": 3
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "results_all = []\n",
    "for path in Path('.').glob('results*/*.csv'):\n",
    "    df = pd.read_csv(path)\n",
    "    df['path'] = str(path)\n",
    "    df['file'] = path.name\n",
    "    df['directory'] = path.parent.name\n",
    "    df = df.groupby(unique_cols, dropna=False).first().reset_index()\n",
    "    results_all.append(df)\n",
    "\n",
    "results_all = pd.concat(results_all)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:31.373052Z",
     "start_time": "2024-02-13T17:08:30.286027Z"
    }
   },
   "id": "e6d8c45c623c1421",
   "execution_count": 4
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "   pretrain_dataset   finetune_dataset finetune_scheme  finetune_subject  \\\n0      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n1      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n2      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n3      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n4      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n..              ...                ...             ...               ...   \n7      BNCI2014_001  Schirrmeister2017        Scheme 3                12   \n8      BNCI2014_001  Schirrmeister2017        Scheme 3                12   \n9      BNCI2014_001  Schirrmeister2017        Scheme 4                12   \n10     BNCI2014_001  Schirrmeister2017        Scheme 4                12   \n11     BNCI2014_001  Schirrmeister2017        Scheme 4                12   \n\n    finetune_fold     directory ref_channel  pretrain_dur  \\\n0               0  results_v0-1     average     50.338649   \n1               0  results_v0-1         NaN     38.459759   \n2               1  results_v0-1     average     50.338649   \n3               1  results_v0-1         NaN     38.459759   \n4               2  results_v0-1     average     50.338649   \n..            ...           ...         ...           ...   \n7               1    results_v2         NaN     38.459759   \n8               2    results_v2         NaN     38.459759   \n9               0    results_v2         NaN     38.459759   \n10              1    results_v2         NaN     38.459759   \n11              2    results_v2         NaN     38.459759   \n\n    finetune_phase1_dur  finetune_phase2_dur  ...  test_size  seed  \\\n0              0.684618             0.209419  ...        160     0   \n1              0.595866             0.205068  ...        160     0   \n2              3.736437             1.792514  ...        160     1   \n3              4.125435             2.843621  ...        160     1   \n4              0.973701             0.627491  ...        160     2   \n..                  ...                  ...  ...        ...   ...   \n7             32.604828             5.074470  ...        160     1   \n8             19.851492             9.291909  ...        160     2   \n9             27.920300             1.994510  ...        160     0   \n10            26.972337             1.918001  ...        160     1   \n11            39.022332             4.568395  ...        160     2   \n\n    pretrain_train_acc  pretrain_valid_acc  finetune_phase1_train_acc  \\\n0             0.543939            0.500963                   0.677083   \n1             0.505195            0.460501                   0.500000   \n2             0.543939            0.500963                   0.785417   \n3             0.505195            0.460501                   0.622917   \n4             0.543939            0.500963                   0.731250   \n..                 ...                 ...                        ...   \n7             0.505195            0.460501                   0.821212   \n8             0.505195            0.460501                   0.822727   \n9             0.505195            0.460501                   0.928788   \n10            0.505195            0.460501                   0.924242   \n11            0.505195            0.460501                   0.931818   \n\n    finetune_phase1_valid_acc  finetune_phase2_valid_acc  test_acc  \\\n0                    0.740458                   0.740458   0.65000   \n1                    0.458015                   0.473282   0.38125   \n2                    0.656489                   0.793893   0.72500   \n3                    0.458015                   0.610687   0.34375   \n4                    0.679389                   0.725191   0.65000   \n..                        ...                        ...       ...   \n7                    0.653409                   0.835227   0.68125   \n8                    0.642045                   0.829545   0.70000   \n9                    0.823864                   0.914773   0.78750   \n10                   0.835227                   0.943182   0.79375   \n11                   0.846591                   0.954545   0.76250   \n\n                                                 path  \\\n0   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n1   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n2   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n3   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n4   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n..                                                ...   \n7   results_v2/pre-BNCI2014_001_fin-Schirrmeister2...   \n8   results_v2/pre-BNCI2014_001_fin-Schirrmeister2...   \n9   results_v2/pre-BNCI2014_001_fin-Schirrmeister2...   \n10  results_v2/pre-BNCI2014_001_fin-Schirrmeister2...   \n11  results_v2/pre-BNCI2014_001_fin-Schirrmeister2...   \n\n                                                 file  \n0    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n1    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n2    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n3    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n4    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n..                                                ...  \n7   pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n8   pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n9   pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n10  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n11  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n\n[17920 rows x 29 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>pretrain_dataset</th>\n      <th>finetune_dataset</th>\n      <th>finetune_scheme</th>\n      <th>finetune_subject</th>\n      <th>finetune_fold</th>\n      <th>directory</th>\n      <th>ref_channel</th>\n      <th>pretrain_dur</th>\n      <th>finetune_phase1_dur</th>\n      <th>finetune_phase2_dur</th>\n      <th>...</th>\n      <th>test_size</th>\n      <th>seed</th>\n      <th>pretrain_train_acc</th>\n      <th>pretrain_valid_acc</th>\n      <th>finetune_phase1_train_acc</th>\n      <th>finetune_phase1_valid_acc</th>\n      <th>finetune_phase2_valid_acc</th>\n      <th>test_acc</th>\n      <th>path</th>\n      <th>file</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>average</td>\n      <td>50.338649</td>\n      <td>0.684618</td>\n      <td>0.209419</td>\n      <td>...</td>\n      <td>160</td>\n      <td>0</td>\n      <td>0.543939</td>\n      <td>0.500963</td>\n      <td>0.677083</td>\n      <td>0.740458</td>\n      <td>0.740458</td>\n      <td>0.65000</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>0.595866</td>\n      <td>0.205068</td>\n      <td>...</td>\n      <td>160</td>\n      <td>0</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.500000</td>\n      <td>0.458015</td>\n      <td>0.473282</td>\n      <td>0.38125</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>1</td>\n      <td>results_v0-1</td>\n      <td>average</td>\n      <td>50.338649</td>\n      <td>3.736437</td>\n      <td>1.792514</td>\n      <td>...</td>\n      <td>160</td>\n      <td>1</td>\n      <td>0.543939</td>\n      <td>0.500963</td>\n      <td>0.785417</td>\n      <td>0.656489</td>\n      <td>0.793893</td>\n      <td>0.72500</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>1</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>4.125435</td>\n      <td>2.843621</td>\n      <td>...</td>\n      <td>160</td>\n      <td>1</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.622917</td>\n      <td>0.458015</td>\n      <td>0.610687</td>\n      <td>0.34375</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>2</td>\n      <td>results_v0-1</td>\n      <td>average</td>\n      <td>50.338649</td>\n      <td>0.973701</td>\n      <td>0.627491</td>\n      <td>...</td>\n      <td>160</td>\n      <td>2</td>\n      <td>0.543939</td>\n      <td>0.500963</td>\n      <td>0.731250</td>\n      <td>0.679389</td>\n      <td>0.725191</td>\n      <td>0.65000</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 3</td>\n      <td>12</td>\n      <td>1</td>\n      <td>results_v2</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>32.604828</td>\n      <td>5.074470</td>\n      <td>...</td>\n      <td>160</td>\n      <td>1</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.821212</td>\n      <td>0.653409</td>\n      <td>0.835227</td>\n      <td>0.68125</td>\n      <td>results_v2/pre-BNCI2014_001_fin-Schirrmeister2...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>8</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 3</td>\n      <td>12</td>\n      <td>2</td>\n      <td>results_v2</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>19.851492</td>\n      <td>9.291909</td>\n      <td>...</td>\n      <td>160</td>\n      <td>2</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.822727</td>\n      <td>0.642045</td>\n      <td>0.829545</td>\n      <td>0.70000</td>\n      <td>results_v2/pre-BNCI2014_001_fin-Schirrmeister2...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>12</td>\n      <td>0</td>\n      <td>results_v2</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>27.920300</td>\n      <td>1.994510</td>\n      <td>...</td>\n      <td>160</td>\n      <td>0</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.928788</td>\n      <td>0.823864</td>\n      <td>0.914773</td>\n      <td>0.78750</td>\n      <td>results_v2/pre-BNCI2014_001_fin-Schirrmeister2...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>10</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>12</td>\n      <td>1</td>\n      <td>results_v2</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>26.972337</td>\n      <td>1.918001</td>\n      <td>...</td>\n      <td>160</td>\n      <td>1</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.924242</td>\n      <td>0.835227</td>\n      <td>0.943182</td>\n      <td>0.79375</td>\n      <td>results_v2/pre-BNCI2014_001_fin-Schirrmeister2...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>11</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>12</td>\n      <td>2</td>\n      <td>results_v2</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>39.022332</td>\n      <td>4.568395</td>\n      <td>...</td>\n      <td>160</td>\n      <td>2</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.931818</td>\n      <td>0.846591</td>\n      <td>0.954545</td>\n      <td>0.76250</td>\n      <td>results_v2/pre-BNCI2014_001_fin-Schirrmeister2...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n  </tbody>\n</table>\n<p>17920 rows × 29 columns</p>\n</div>"
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_all"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:31.381866Z",
     "start_time": "2024-02-13T17:08:31.357196Z"
    }
   },
   "id": "1752732e388b5279",
   "execution_count": 5
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "   pretrain_dataset   finetune_dataset finetune_scheme  finetune_subject  \\\n1      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n3      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n5      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n7      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n9      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n..              ...                ...             ...               ...   \n95     BNCI2014_001  Schirrmeister2017        Scheme 4                12   \n96     BNCI2014_001  Schirrmeister2017        Scheme 4                12   \n97     BNCI2014_001  Schirrmeister2017        Scheme 4                12   \n98     BNCI2014_001  Schirrmeister2017        Scheme 4                12   \n99     BNCI2014_001  Schirrmeister2017        Scheme 4                12   \n\n    finetune_fold     directory ref_channel  pretrain_dur  \\\n1               0  results_v0-1         NaN     38.459759   \n3               1  results_v0-1         NaN     38.459759   \n5               2  results_v0-1         NaN     38.459759   \n7               3  results_v0-1         NaN     38.459759   \n9               4  results_v0-1         NaN     38.459759   \n..            ...           ...         ...           ...   \n95             15  results_v0-1         NaN     38.459759   \n96             16  results_v0-1         NaN     38.459759   \n97             17  results_v0-1         NaN     38.459759   \n98             18  results_v0-1         NaN     38.459759   \n99             19  results_v0-1         NaN     38.459759   \n\n    finetune_phase1_dur  finetune_phase2_dur  ...  test_size  seed  \\\n1              0.595866             0.205068  ...        160     0   \n3              4.125435             2.843621  ...        160     1   \n5              0.748193             0.685473  ...        160     2   \n7              0.165159             0.035121  ...        160     3   \n9              2.531078             2.194521  ...        160     4   \n..                  ...                  ...  ...        ...   ...   \n95            22.591941             2.348470  ...        160    15   \n96            10.997515             0.834382  ...        160    16   \n97            16.140081             0.776989  ...        160    17   \n98            14.286297             1.922483  ...        160    18   \n99            16.005492             2.028342  ...        160    19   \n\n    pretrain_train_acc  pretrain_valid_acc  finetune_phase1_train_acc  \\\n1             0.505195            0.460501                   0.500000   \n3             0.505195            0.460501                   0.622917   \n5             0.505195            0.460501                   0.533333   \n7             0.505195            0.460501                   0.368750   \n9             0.505195            0.460501                   0.610417   \n..                 ...                 ...                        ...   \n95            0.505195            0.460501                   0.890909   \n96            0.505195            0.460501                   0.850000   \n97            0.505195            0.460501                   0.865152   \n98            0.505195            0.460501                   0.857576   \n99            0.505195            0.460501                   0.895455   \n\n    finetune_phase1_valid_acc  finetune_phase2_valid_acc  test_acc  \\\n1                    0.458015                   0.473282   0.38125   \n3                    0.458015                   0.610687   0.34375   \n5                    0.419847                   0.496183   0.41250   \n7                    0.511450                   0.511450   0.38125   \n9                    0.442748                   0.625954   0.35625   \n..                        ...                        ...       ...   \n95                   0.829545                   0.926136   0.81875   \n96                   0.840909                   0.875000   0.77500   \n97                   0.857955                   0.846591   0.74375   \n98                   0.835227                   0.886364   0.76250   \n99                   0.835227                   0.897727   0.76250   \n\n                                                 path  \\\n1   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n3   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n5   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n7   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n9   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n..                                                ...   \n95  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n96  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n97  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n98  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n99  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n\n                                                 file  \n1    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n3    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n5    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n7    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n9    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n..                                                ...  \n95  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n96  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n97  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n98  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n99  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n\n[12800 rows x 29 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>pretrain_dataset</th>\n      <th>finetune_dataset</th>\n      <th>finetune_scheme</th>\n      <th>finetune_subject</th>\n      <th>finetune_fold</th>\n      <th>directory</th>\n      <th>ref_channel</th>\n      <th>pretrain_dur</th>\n      <th>finetune_phase1_dur</th>\n      <th>finetune_phase2_dur</th>\n      <th>...</th>\n      <th>test_size</th>\n      <th>seed</th>\n      <th>pretrain_train_acc</th>\n      <th>pretrain_valid_acc</th>\n      <th>finetune_phase1_train_acc</th>\n      <th>finetune_phase1_valid_acc</th>\n      <th>finetune_phase2_valid_acc</th>\n      <th>test_acc</th>\n      <th>path</th>\n      <th>file</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>1</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>0.595866</td>\n      <td>0.205068</td>\n      <td>...</td>\n      <td>160</td>\n      <td>0</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.500000</td>\n      <td>0.458015</td>\n      <td>0.473282</td>\n      <td>0.38125</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>1</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>4.125435</td>\n      <td>2.843621</td>\n      <td>...</td>\n      <td>160</td>\n      <td>1</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.622917</td>\n      <td>0.458015</td>\n      <td>0.610687</td>\n      <td>0.34375</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>2</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>0.748193</td>\n      <td>0.685473</td>\n      <td>...</td>\n      <td>160</td>\n      <td>2</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.533333</td>\n      <td>0.419847</td>\n      <td>0.496183</td>\n      <td>0.41250</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>7</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>3</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>0.165159</td>\n      <td>0.035121</td>\n      <td>...</td>\n      <td>160</td>\n      <td>3</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.368750</td>\n      <td>0.511450</td>\n      <td>0.511450</td>\n      <td>0.38125</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>9</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>4</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>2.531078</td>\n      <td>2.194521</td>\n      <td>...</td>\n      <td>160</td>\n      <td>4</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.610417</td>\n      <td>0.442748</td>\n      <td>0.625954</td>\n      <td>0.35625</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>95</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>12</td>\n      <td>15</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>22.591941</td>\n      <td>2.348470</td>\n      <td>...</td>\n      <td>160</td>\n      <td>15</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.890909</td>\n      <td>0.829545</td>\n      <td>0.926136</td>\n      <td>0.81875</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>96</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>12</td>\n      <td>16</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>10.997515</td>\n      <td>0.834382</td>\n      <td>...</td>\n      <td>160</td>\n      <td>16</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.850000</td>\n      <td>0.840909</td>\n      <td>0.875000</td>\n      <td>0.77500</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>97</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>12</td>\n      <td>17</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>16.140081</td>\n      <td>0.776989</td>\n      <td>...</td>\n      <td>160</td>\n      <td>17</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.865152</td>\n      <td>0.857955</td>\n      <td>0.846591</td>\n      <td>0.74375</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>98</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>12</td>\n      <td>18</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>14.286297</td>\n      <td>1.922483</td>\n      <td>...</td>\n      <td>160</td>\n      <td>18</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.857576</td>\n      <td>0.835227</td>\n      <td>0.886364</td>\n      <td>0.76250</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>99</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>12</td>\n      <td>19</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>38.459759</td>\n      <td>16.005492</td>\n      <td>2.028342</td>\n      <td>...</td>\n      <td>160</td>\n      <td>19</td>\n      <td>0.505195</td>\n      <td>0.460501</td>\n      <td>0.895455</td>\n      <td>0.835227</td>\n      <td>0.897727</td>\n      <td>0.76250</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n  </tbody>\n</table>\n<p>12800 rows × 29 columns</p>\n</div>"
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Filter results\n",
    "conditions = dict(\n",
    "    ref_channel=None,\n",
    "    # ref_channel='average',\n",
    "    # directory='results_v2',\n",
    "    directory='results_v0-1',\n",
    ")\n",
    "\n",
    "results = results_all.copy()\n",
    "for k, v in conditions.items():\n",
    "    if v is not None:\n",
    "        mask = results[k] == v\n",
    "    else:\n",
    "        mask = results[k].isna()\n",
    "    results = results[mask]\n",
    "\n",
    "results"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:31.421787Z",
     "start_time": "2024-02-13T17:08:31.377073Z"
    }
   },
   "id": "c72e414890b3530e",
   "execution_count": 6
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "pretrain_dataset   finetune_dataset \nBNCI2014_001       BNCI2014_001          720\n                   BNCI2014_004          720\n                   Schirrmeister2017    1120\nBNCI2014_004       BNCI2014_001          720\n                   BNCI2014_004          720\n                   Schirrmeister2017    1120\nLee2019_MI         BNCI2014_001          720\n                   BNCI2014_004          720\n                   Schirrmeister2017    1120\nPhysionetMI        BNCI2014_001          720\n                   BNCI2014_004          720\n                   Schirrmeister2017    1120\nSchirrmeister2017  BNCI2014_001          720\n                   BNCI2014_004          720\n                   Schirrmeister2017    1120\nName: test_acc, dtype: int64"
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# check the number of results\n",
    "results.groupby(\n",
    "    # ['pretrain_dataset', 'finetune_dataset', 'finetune_subject']\n",
    "    ['pretrain_dataset', 'finetune_dataset']  # n_subjects * n folds * 4 schemes\n",
    ").test_acc.count()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:31.423339Z",
     "start_time": "2024-02-13T17:08:31.400501Z"
    }
   },
   "id": "73947d8454a91d26",
   "execution_count": 7
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "# check that all experiments logged the same metrics for the same pretraining phase\n",
    "assert (results.groupby('pretrain_dataset')[[\n",
    "    'pretrain_dur',\n",
    "    'pretrain_n_epochs',\n",
    "    'pretrain_train_size',\n",
    "    'pretrain_valid_size',\n",
    "    'pretrain_dataset',\n",
    "    'pretrain_train_acc',\n",
    "    'pretrain_valid_acc',\n",
    "]].nunique() == 1).all().all()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:31.464511Z",
     "start_time": "2024-02-13T17:08:31.406206Z"
    }
   },
   "id": "e7d7b97d0c632fd5",
   "execution_count": 8
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "# Check that all folds were executed once only each\n",
    "assert results.groupby('finetune_fold').test_acc.count().nunique() == 1"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:31.830212Z",
     "start_time": "2024-02-13T17:08:31.695352Z"
    }
   },
   "id": "339204d9f056df57",
   "execution_count": 9
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "# Better check that all folds were executed once only each\n",
    "assert (results.groupby(\n",
    "    ['pretrain_dataset', 'finetune_dataset', 'finetune_scheme', 'finetune_subject',\n",
    "     'finetune_fold']).test_acc.count() == 1).all()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:31.941695Z",
     "start_time": "2024-02-13T17:08:31.912070Z"
    }
   },
   "id": "468d87fb6861c4b7",
   "execution_count": 10
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "Empty DataFrame\nColumns: [pretrain_dur, finetune_phase1_dur, finetune_phase2_dur, pretrain_n_epochs, finetune_phase1_n_epochs, finetune_phase2_n_epochs, pretrain_train_size, pretrain_valid_size, finetune_phase1_train_size, finetune_phase1_valid_size, finetune_phase2_train_size, finetune_phase2_valid_size, test_size, seed, pretrain_train_acc, pretrain_valid_acc, finetune_phase1_train_acc, finetune_phase1_valid_acc, finetune_phase2_valid_acc, test_acc, path, file]\nIndex: []\n\n[0 rows x 22 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></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th>pretrain_dur</th>\n      <th>finetune_phase1_dur</th>\n      <th>finetune_phase2_dur</th>\n      <th>pretrain_n_epochs</th>\n      <th>finetune_phase1_n_epochs</th>\n      <th>finetune_phase2_n_epochs</th>\n      <th>pretrain_train_size</th>\n      <th>pretrain_valid_size</th>\n      <th>finetune_phase1_train_size</th>\n      <th>finetune_phase1_valid_size</th>\n      <th>...</th>\n      <th>test_size</th>\n      <th>seed</th>\n      <th>pretrain_train_acc</th>\n      <th>pretrain_valid_acc</th>\n      <th>finetune_phase1_train_acc</th>\n      <th>finetune_phase1_valid_acc</th>\n      <th>finetune_phase2_valid_acc</th>\n      <th>test_acc</th>\n      <th>path</th>\n      <th>file</th>\n    </tr>\n    <tr>\n      <th>pretrain_dataset</th>\n      <th>finetune_dataset</th>\n      <th>finetune_scheme</th>\n      <th>finetune_subject</th>\n      <th>finetune_fold</th>\n      <th>directory</th>\n      <th>ref_channel</th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n  </tbody>\n</table>\n<p>0 rows × 22 columns</p>\n</div>"
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "c = results.groupby(unique_cols).test_acc.count()\n",
    "results.set_index(unique_cols, inplace=False).filter(c[c != 1].index, axis='index')"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:32.229191Z",
     "start_time": "2024-02-13T17:08:32.205542Z"
    }
   },
   "id": "20d2cb7d4d33ecd5",
   "execution_count": 11
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "   pretrain_dataset   finetune_dataset finetune_scheme  finetune_subject  \\\n1      BNCI2014_001  Schirrmeister2017        Scheme 1                 8   \n26     BNCI2014_001  Schirrmeister2017        Scheme 2                 8   \n51     BNCI2014_001  Schirrmeister2017        Scheme 3                 8   \n76     BNCI2014_001  Schirrmeister2017        Scheme 4                 8   \n1      BNCI2014_001       BNCI2014_004        Scheme 1                 2   \n..              ...                ...             ...               ...   \n76     BNCI2014_001  Schirrmeister2017        Scheme 4                 5   \n1      BNCI2014_001  Schirrmeister2017        Scheme 1                12   \n26     BNCI2014_001  Schirrmeister2017        Scheme 2                12   \n51     BNCI2014_001  Schirrmeister2017        Scheme 3                12   \n76     BNCI2014_001  Schirrmeister2017        Scheme 4                12   \n\n    finetune_fold     directory ref_channel  pretrain_n_epochs  \\\n1               0  results_v0-1         NaN                 60   \n26              0  results_v0-1         NaN                 60   \n51              0  results_v0-1         NaN                 60   \n76              0  results_v0-1         NaN                 60   \n1               0  results_v0-1         NaN                 60   \n..            ...           ...         ...                ...   \n76              0  results_v0-1         NaN                 60   \n1               0  results_v0-1         NaN                 60   \n26              0  results_v0-1         NaN                 60   \n51              0  results_v0-1         NaN                 60   \n76              0  results_v0-1         NaN                 60   \n\n    finetune_phase1_n_epochs  finetune_phase2_n_epochs  pretrain_train_size  \\\n1                         16                        21                 4620   \n26                        60                        61                 4620   \n51                        61                        62                 4620   \n76                       163                       181                 4620   \n1                         29                        64                 4620   \n..                       ...                       ...                  ...   \n76                        85                        93                 4620   \n1                        162                       193                 4620   \n26                       165                       167                 4620   \n51                       411                       447                 4620   \n76                       430                       450                 4620   \n\n    pretrain_valid_size  finetune_phase1_train_size  \\\n1                   519                         480   \n26                  519                         480   \n51                  519                         480   \n76                  519                         480   \n1                   519                         300   \n..                  ...                         ...   \n76                  519                         540   \n1                   519                         660   \n26                  519                         660   \n51                  519                         660   \n76                  519                         660   \n\n    finetune_phase1_valid_size  finetune_phase2_train_size  \\\n1                          131                         600   \n26                         131                         600   \n51                         131                         600   \n76                         131                         600   \n1                           80                         360   \n..                         ...                         ...   \n76                         144                         720   \n1                          176                         840   \n26                         176                         840   \n51                         176                         840   \n76                         176                         840   \n\n    finetune_phase2_valid_size  test_size  seed  \\\n1                          131        160     0   \n26                         131        160     0   \n51                         131        160     0   \n76                         131        160     0   \n1                           80        280     0   \n..                         ...        ...   ...   \n76                         144        160     0   \n1                          176        160     0   \n26                         176        160     0   \n51                         176        160     0   \n76                         176        160     0   \n\n                                                 path  \\\n1   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n26  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n51  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n76  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n1   results_v0-1/pre-BNCI2014_001_fin-BNCI2014_004...   \n..                                                ...   \n76  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n1   results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n26  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n51  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n76  results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...   \n\n                                                 file  \n1    pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n26   pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n51   pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n76   pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv  \n1         pre-BNCI2014_001_fin-BNCI2014_004_sub-2.csv  \n..                                                ...  \n76   pre-BNCI2014_001_fin-Schirrmeister2017_sub-5.csv  \n1   pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n26  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n51  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n76  pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv  \n\n[640 rows x 20 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>pretrain_dataset</th>\n      <th>finetune_dataset</th>\n      <th>finetune_scheme</th>\n      <th>finetune_subject</th>\n      <th>finetune_fold</th>\n      <th>directory</th>\n      <th>ref_channel</th>\n      <th>pretrain_n_epochs</th>\n      <th>finetune_phase1_n_epochs</th>\n      <th>finetune_phase2_n_epochs</th>\n      <th>pretrain_train_size</th>\n      <th>pretrain_valid_size</th>\n      <th>finetune_phase1_train_size</th>\n      <th>finetune_phase1_valid_size</th>\n      <th>finetune_phase2_train_size</th>\n      <th>finetune_phase2_valid_size</th>\n      <th>test_size</th>\n      <th>seed</th>\n      <th>path</th>\n      <th>file</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>1</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>8</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>16</td>\n      <td>21</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>480</td>\n      <td>131</td>\n      <td>600</td>\n      <td>131</td>\n      <td>160</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>26</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 2</td>\n      <td>8</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>60</td>\n      <td>61</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>480</td>\n      <td>131</td>\n      <td>600</td>\n      <td>131</td>\n      <td>160</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>51</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 3</td>\n      <td>8</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>61</td>\n      <td>62</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>480</td>\n      <td>131</td>\n      <td>600</td>\n      <td>131</td>\n      <td>160</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>76</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>8</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>163</td>\n      <td>181</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>480</td>\n      <td>131</td>\n      <td>600</td>\n      <td>131</td>\n      <td>160</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-8.csv</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>BNCI2014_001</td>\n      <td>BNCI2014_004</td>\n      <td>Scheme 1</td>\n      <td>2</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>29</td>\n      <td>64</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>300</td>\n      <td>80</td>\n      <td>360</td>\n      <td>80</td>\n      <td>280</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-BNCI2014_004...</td>\n      <td>pre-BNCI2014_001_fin-BNCI2014_004_sub-2.csv</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th>76</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>5</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>85</td>\n      <td>93</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>540</td>\n      <td>144</td>\n      <td>720</td>\n      <td>144</td>\n      <td>160</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-5.csv</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 1</td>\n      <td>12</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>162</td>\n      <td>193</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>660</td>\n      <td>176</td>\n      <td>840</td>\n      <td>176</td>\n      <td>160</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>26</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 2</td>\n      <td>12</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>165</td>\n      <td>167</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>660</td>\n      <td>176</td>\n      <td>840</td>\n      <td>176</td>\n      <td>160</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>51</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 3</td>\n      <td>12</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>411</td>\n      <td>447</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>660</td>\n      <td>176</td>\n      <td>840</td>\n      <td>176</td>\n      <td>160</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n    <tr>\n      <th>76</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>Scheme 4</td>\n      <td>12</td>\n      <td>0</td>\n      <td>results_v0-1</td>\n      <td>NaN</td>\n      <td>60</td>\n      <td>430</td>\n      <td>450</td>\n      <td>4620</td>\n      <td>519</td>\n      <td>660</td>\n      <td>176</td>\n      <td>840</td>\n      <td>176</td>\n      <td>160</td>\n      <td>0</td>\n      <td>results_v0-1/pre-BNCI2014_001_fin-Schirrmeiste...</td>\n      <td>pre-BNCI2014_001_fin-Schirrmeister2017_sub-12.csv</td>\n    </tr>\n  </tbody>\n</table>\n<p>640 rows × 20 columns</p>\n</div>"
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results[results.finetune_fold == 0][results.columns[results.dtypes != 'float']]"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:32.487585Z",
     "start_time": "2024-02-13T17:08:32.416072Z"
    }
   },
   "id": "89021e165822c64b",
   "execution_count": 12
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "                  finetune_phase1_train_size                     \\\n                                      unique   mean         std   \nfinetune_dataset                                                  \nBNCI2014_001                           [180]  180.0    0.000000   \nBNCI2014_004                           [300]  300.0    0.000000   \nSchirrmeister2017  [480, 660, 600, 240, 540]  600.0  113.399467   \n\n                           finetune_phase1_valid_size                         \\\n                                               unique        mean        std   \nfinetune_dataset                                                               \nBNCI2014_001                                     [58]   58.000000   0.000000   \nBNCI2014_004                             [80, 84, 88]   81.777778   2.740120   \nSchirrmeister2017  [131, 176, 160, 163, 64, 144, 180]  160.714286  30.195951   \n\n                    test_size                         \n                       unique        mean        std  \nfinetune_dataset                                      \nBNCI2014_001            [288]  288.000000   0.000000  \nBNCI2014_004       [280, 320]  315.555556  12.572534  \nSchirrmeister2017       [160]  160.000000   0.000000  ",
      "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 tr th {\n        text-align: left;\n    }\n\n    .dataframe thead tr:last-of-type th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr>\n      <th></th>\n      <th colspan=\"3\" halign=\"left\">finetune_phase1_train_size</th>\n      <th colspan=\"3\" halign=\"left\">finetune_phase1_valid_size</th>\n      <th colspan=\"3\" halign=\"left\">test_size</th>\n    </tr>\n    <tr>\n      <th></th>\n      <th>unique</th>\n      <th>mean</th>\n      <th>std</th>\n      <th>unique</th>\n      <th>mean</th>\n      <th>std</th>\n      <th>unique</th>\n      <th>mean</th>\n      <th>std</th>\n    </tr>\n    <tr>\n      <th>finetune_dataset</th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>BNCI2014_001</th>\n      <td>[180]</td>\n      <td>180.0</td>\n      <td>0.000000</td>\n      <td>[58]</td>\n      <td>58.000000</td>\n      <td>0.000000</td>\n      <td>[288]</td>\n      <td>288.000000</td>\n      <td>0.000000</td>\n    </tr>\n    <tr>\n      <th>BNCI2014_004</th>\n      <td>[300]</td>\n      <td>300.0</td>\n      <td>0.000000</td>\n      <td>[80, 84, 88]</td>\n      <td>81.777778</td>\n      <td>2.740120</td>\n      <td>[280, 320]</td>\n      <td>315.555556</td>\n      <td>12.572534</td>\n    </tr>\n    <tr>\n      <th>Schirrmeister2017</th>\n      <td>[480, 660, 600, 240, 540]</td>\n      <td>600.0</td>\n      <td>113.399467</td>\n      <td>[131, 176, 160, 163, 64, 144, 180]</td>\n      <td>160.714286</td>\n      <td>30.195951</td>\n      <td>[160]</td>\n      <td>160.000000</td>\n      <td>0.000000</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 126,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# check amount of data in the different subsets\n",
    "# inconsistent with Xie2023 \n",
    "# --> because of drop_last=True for train sets! \n",
    "# But actually it's not a problem because the train set is reshuffled at each epoch\n",
    "\n",
    "results.groupby('finetune_dataset')[\n",
    "    ['finetune_phase1_train_size', 'finetune_phase1_valid_size', 'test_size']].agg(\n",
    "    ['unique', 'mean', 'std'])  #'min', 'max', "
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-16T04:28:21.813554Z",
     "start_time": "2024-02-16T04:28:21.747492Z"
    }
   },
   "id": "8d8950c334cb5387",
   "execution_count": 126
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Durations of the experiment phases"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "fe346480553ad560"
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "pretrain_dataset\nBNCI2014_001          38.459759\nBNCI2014_004          58.810894\nLee2019_MI            90.338237\nPhysionetMI          846.469128\nSchirrmeister2017    474.220997\nName: pretrain_dur, dtype: float64"
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Pre-training\n",
    "results.groupby('pretrain_dataset').first().pretrain_dur"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:33.627989Z",
     "start_time": "2024-02-13T17:08:33.622563Z"
    }
   },
   "id": "c8c14a42917b8ce6",
   "execution_count": 14
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "                                  finetune_phase1_dur            \\\n                                                 mean       std   \nfinetune_dataset  finetune_scheme                                 \nBNCI2014_001      Scheme 1                   0.710596  0.614789   \n                  Scheme 2                   1.704826  1.202186   \n                  Scheme 3                   1.384863  0.927094   \n                  Scheme 4                   1.769336  1.235909   \nBNCI2014_004      Scheme 1                   0.970478  0.958157   \n                  Scheme 2                   2.027992  1.659104   \n                  Scheme 3                   1.792085  1.504715   \n                  Scheme 4                   1.849485  1.412250   \nSchirrmeister2017 Scheme 1                   2.571169  2.625265   \n                  Scheme 2                   5.779999  4.365365   \n                  Scheme 3                   4.825514  3.722407   \n                  Scheme 4                   6.852640  5.164285   \n\n                                  finetune_phase2_dur            \n                                                 mean       std  \nfinetune_dataset  finetune_scheme                                \nBNCI2014_001      Scheme 1                   0.517394  0.493343  \n                  Scheme 2                   0.818922  0.543866  \n                  Scheme 3                   0.711224  0.497875  \n                  Scheme 4                   0.820012  0.548374  \nBNCI2014_004      Scheme 1                   0.539528  0.625279  \n                  Scheme 2                   0.753290  0.825675  \n                  Scheme 3                   0.666671  0.691999  \n                  Scheme 4                   0.715240  0.729179  \nSchirrmeister2017 Scheme 1                   1.319154  6.934650  \n                  Scheme 2                   1.510370  1.340723  \n                  Scheme 3                   1.474296  1.563769  \n                  Scheme 4                   1.574592  1.265293  ",
      "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 tr th {\n        text-align: left;\n    }\n\n    .dataframe thead tr:last-of-type th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr>\n      <th></th>\n      <th></th>\n      <th colspan=\"2\" halign=\"left\">finetune_phase1_dur</th>\n      <th colspan=\"2\" halign=\"left\">finetune_phase2_dur</th>\n    </tr>\n    <tr>\n      <th></th>\n      <th></th>\n      <th>mean</th>\n      <th>std</th>\n      <th>mean</th>\n      <th>std</th>\n    </tr>\n    <tr>\n      <th>finetune_dataset</th>\n      <th>finetune_scheme</th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">BNCI2014_001</th>\n      <th>Scheme 1</th>\n      <td>0.710596</td>\n      <td>0.614789</td>\n      <td>0.517394</td>\n      <td>0.493343</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>1.704826</td>\n      <td>1.202186</td>\n      <td>0.818922</td>\n      <td>0.543866</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>1.384863</td>\n      <td>0.927094</td>\n      <td>0.711224</td>\n      <td>0.497875</td>\n    </tr>\n    <tr>\n      <th>Scheme 4</th>\n      <td>1.769336</td>\n      <td>1.235909</td>\n      <td>0.820012</td>\n      <td>0.548374</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">BNCI2014_004</th>\n      <th>Scheme 1</th>\n      <td>0.970478</td>\n      <td>0.958157</td>\n      <td>0.539528</td>\n      <td>0.625279</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>2.027992</td>\n      <td>1.659104</td>\n      <td>0.753290</td>\n      <td>0.825675</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>1.792085</td>\n      <td>1.504715</td>\n      <td>0.666671</td>\n      <td>0.691999</td>\n    </tr>\n    <tr>\n      <th>Scheme 4</th>\n      <td>1.849485</td>\n      <td>1.412250</td>\n      <td>0.715240</td>\n      <td>0.729179</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">Schirrmeister2017</th>\n      <th>Scheme 1</th>\n      <td>2.571169</td>\n      <td>2.625265</td>\n      <td>1.319154</td>\n      <td>6.934650</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>5.779999</td>\n      <td>4.365365</td>\n      <td>1.510370</td>\n      <td>1.340723</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>4.825514</td>\n      <td>3.722407</td>\n      <td>1.474296</td>\n      <td>1.563769</td>\n    </tr>\n    <tr>\n      <th>Scheme 4</th>\n      <td>6.852640</td>\n      <td>5.164285</td>\n      <td>1.574592</td>\n      <td>1.265293</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fine-tuning\n",
    "# finetuning longer on Schirrmeister\n",
    "results.groupby(['finetune_dataset', 'finetune_scheme'])[['finetune_phase1_dur', 'finetune_phase2_dur']].agg(\n",
    "    ['mean', 'std'])"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:34.114863Z",
     "start_time": "2024-02-13T17:08:34.102732Z"
    }
   },
   "id": "eae0b8da07c265af",
   "execution_count": 15
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Scores\n",
    "### Within one commit\n",
    "#### Average"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "e68154c3b6467cfe"
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "                                                         mean       std  count\nfinetune_dataset  finetune_scheme pretrain_dataset                            \nBNCI2014_001      Scheme 4        BNCI2014_001       0.518866  0.113234    180\n                                  BNCI2014_004       0.496123  0.127056    180\n                                  Lee2019_MI         0.495390  0.126802    180\n                                  PhysionetMI        0.525463  0.118771    180\n                                  Schirrmeister2017  0.448052  0.113954    180\nBNCI2014_004      Scheme 4        BNCI2014_001       0.763973  0.145494    180\n                                  BNCI2014_004       0.779829  0.143288    180\n                                  Lee2019_MI         0.778242  0.145587    180\n                                  PhysionetMI        0.764908  0.139890    180\n                                  Schirrmeister2017  0.768765  0.147409    180\nSchirrmeister2017 Scheme 4        BNCI2014_001       0.644531  0.095613    280\n                                  BNCI2014_004       0.640379  0.089635    280\n                                  Lee2019_MI         0.642455  0.090689    280\n                                  PhysionetMI        0.649687  0.094683    280\n                                  Schirrmeister2017  0.684531  0.077068    280",
      "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></th>\n      <th></th>\n      <th>mean</th>\n      <th>std</th>\n      <th>count</th>\n    </tr>\n    <tr>\n      <th>finetune_dataset</th>\n      <th>finetune_scheme</th>\n      <th>pretrain_dataset</th>\n      <th></th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th rowspan=\"5\" valign=\"top\">BNCI2014_001</th>\n      <th rowspan=\"5\" valign=\"top\">Scheme 4</th>\n      <th>BNCI2014_001</th>\n      <td>0.518866</td>\n      <td>0.113234</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th>BNCI2014_004</th>\n      <td>0.496123</td>\n      <td>0.127056</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th>Lee2019_MI</th>\n      <td>0.495390</td>\n      <td>0.126802</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th>PhysionetMI</th>\n      <td>0.525463</td>\n      <td>0.118771</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th>Schirrmeister2017</th>\n      <td>0.448052</td>\n      <td>0.113954</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th rowspan=\"5\" valign=\"top\">BNCI2014_004</th>\n      <th rowspan=\"5\" valign=\"top\">Scheme 4</th>\n      <th>BNCI2014_001</th>\n      <td>0.763973</td>\n      <td>0.145494</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th>BNCI2014_004</th>\n      <td>0.779829</td>\n      <td>0.143288</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th>Lee2019_MI</th>\n      <td>0.778242</td>\n      <td>0.145587</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th>PhysionetMI</th>\n      <td>0.764908</td>\n      <td>0.139890</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th>Schirrmeister2017</th>\n      <td>0.768765</td>\n      <td>0.147409</td>\n      <td>180</td>\n    </tr>\n    <tr>\n      <th rowspan=\"5\" valign=\"top\">Schirrmeister2017</th>\n      <th rowspan=\"5\" valign=\"top\">Scheme 4</th>\n      <th>BNCI2014_001</th>\n      <td>0.644531</td>\n      <td>0.095613</td>\n      <td>280</td>\n    </tr>\n    <tr>\n      <th>BNCI2014_004</th>\n      <td>0.640379</td>\n      <td>0.089635</td>\n      <td>280</td>\n    </tr>\n    <tr>\n      <th>Lee2019_MI</th>\n      <td>0.642455</td>\n      <td>0.090689</td>\n      <td>280</td>\n    </tr>\n    <tr>\n      <th>PhysionetMI</th>\n      <td>0.649687</td>\n      <td>0.094683</td>\n      <td>280</td>\n    </tr>\n    <tr>\n      <th>Schirrmeister2017</th>\n      <td>0.684531</td>\n      <td>0.077068</td>\n      <td>280</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# main test results\n",
    "results[results.finetune_scheme == 'Scheme 4'].groupby(\n",
    "    ['finetune_dataset', 'finetune_scheme', 'pretrain_dataset']).test_acc.agg(['mean', 'std', 'count'])"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:35.348234Z",
     "start_time": "2024-02-13T17:08:35.334610Z"
    }
   },
   "id": "9dc7aba15f00f9ac",
   "execution_count": 16
  },
  {
   "cell_type": "markdown",
   "source": [
    "#### Rank"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "6ab727d8abd309cf"
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "                                         rank  test_acc\nfinetune_dataset  pretrain_dataset                     \nBNCI2014_001      PhysionetMI        1.750000  0.525463\n                  Lee2019_MI         2.461111  0.495390\n                  BNCI2014_004       2.483333  0.496123\n                  Schirrmeister2017  3.305556  0.448052\nBNCI2014_004      Lee2019_MI         1.983333  0.778242\n                  Schirrmeister2017  2.608333  0.768765\n                  BNCI2014_001       2.686111  0.763973\n                  PhysionetMI        2.722222  0.764908\nSchirrmeister2017 PhysionetMI        2.357143  0.649687\n                  Lee2019_MI         2.453571  0.642455\n                  BNCI2014_001       2.530357  0.644531\n                  BNCI2014_004       2.658929  0.640379",
      "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></th>\n      <th>rank</th>\n      <th>test_acc</th>\n    </tr>\n    <tr>\n      <th>finetune_dataset</th>\n      <th>pretrain_dataset</th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">BNCI2014_001</th>\n      <th>PhysionetMI</th>\n      <td>1.750000</td>\n      <td>0.525463</td>\n    </tr>\n    <tr>\n      <th>Lee2019_MI</th>\n      <td>2.461111</td>\n      <td>0.495390</td>\n    </tr>\n    <tr>\n      <th>BNCI2014_004</th>\n      <td>2.483333</td>\n      <td>0.496123</td>\n    </tr>\n    <tr>\n      <th>Schirrmeister2017</th>\n      <td>3.305556</td>\n      <td>0.448052</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">BNCI2014_004</th>\n      <th>Lee2019_MI</th>\n      <td>1.983333</td>\n      <td>0.778242</td>\n    </tr>\n    <tr>\n      <th>Schirrmeister2017</th>\n      <td>2.608333</td>\n      <td>0.768765</td>\n    </tr>\n    <tr>\n      <th>BNCI2014_001</th>\n      <td>2.686111</td>\n      <td>0.763973</td>\n    </tr>\n    <tr>\n      <th>PhysionetMI</th>\n      <td>2.722222</td>\n      <td>0.764908</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">Schirrmeister2017</th>\n      <th>PhysionetMI</th>\n      <td>2.357143</td>\n      <td>0.649687</td>\n    </tr>\n    <tr>\n      <th>Lee2019_MI</th>\n      <td>2.453571</td>\n      <td>0.642455</td>\n    </tr>\n    <tr>\n      <th>BNCI2014_001</th>\n      <td>2.530357</td>\n      <td>0.644531</td>\n    </tr>\n    <tr>\n      <th>BNCI2014_004</th>\n      <td>2.658929</td>\n      <td>0.640379</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Rank pre-training datasets\n",
    "cols_rank_pretrain_dataset = [c for c in unique_cols if c != 'pretrain_dataset']\n",
    "d = results.copy()[results.finetune_scheme == 'Scheme 4']\n",
    "d = d[d.pretrain_dataset.ne(d.finetune_dataset)]\n",
    "d['rank'] = d.groupby(cols_rank_pretrain_dataset, dropna=False).test_acc.rank(ascending=False, na_option='keep')\n",
    "d.groupby(['finetune_dataset', 'pretrain_dataset']).agg({'rank': 'mean', 'test_acc': 'mean'}).sort_values(\n",
    "    'rank').sort_index(level=0, sort_remaining=False)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:36.764858Z",
     "start_time": "2024-02-13T17:08:36.748033Z"
    }
   },
   "id": "41517747bdcd3527",
   "execution_count": 17
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "                                                         rank  test_acc\nfinetune_dataset  pretrain_dataset  finetune_scheme                    \nBNCI2014_001      BNCI2014_004      Scheme 4         1.469444  0.496123\n                                    Scheme 2         2.041667  0.470120\n                                    Scheme 3         2.691667  0.448457\n                                    Scheme 1         3.797222  0.395197\n                  Lee2019_MI        Scheme 4         1.602778  0.495390\n                                    Scheme 2         2.138889  0.479398\n                                    Scheme 3         2.513889  0.466667\n                                    Scheme 1         3.744444  0.402836\n                  PhysionetMI       Scheme 4         1.816667  0.525463\n                                    Scheme 2         2.072222  0.519522\n                                    Scheme 3         2.502778  0.507523\n                                    Scheme 1         3.608333  0.458063\n                  Schirrmeister2017 Scheme 4         1.688889  0.448052\n                                    Scheme 2         1.886111  0.444541\n                                    Scheme 3         2.886111  0.399711\n                                    Scheme 1         3.538889  0.376485\nBNCI2014_004      BNCI2014_001      Scheme 4         1.638889  0.763973\n                                    Scheme 2         2.013889  0.752059\n                                    Scheme 3         2.658333  0.725010\n                                    Scheme 1         3.688889  0.674928\n                  Lee2019_MI        Scheme 4         1.702778  0.778242\n                                    Scheme 2         2.066667  0.762879\n                                    Scheme 3         2.641667  0.751200\n                                    Scheme 1         3.588889  0.712314\n                  PhysionetMI       Scheme 4         1.586111  0.764908\n                                    Scheme 2         1.927778  0.754906\n                                    Scheme 3         2.836111  0.726131\n                                    Scheme 1         3.650000  0.695937\n                  Schirrmeister2017 Scheme 4         1.594444  0.768765\n                                    Scheme 2         2.011111  0.754866\n                                    Scheme 3         2.633333  0.725719\n                                    Scheme 1         3.761111  0.680340\nSchirrmeister2017 BNCI2014_001      Scheme 4         1.367857  0.644531\n                                    Scheme 2         1.841071  0.622076\n                                    Scheme 3         2.860714  0.571027\n                                    Scheme 1         3.930357  0.468929\n                  BNCI2014_004      Scheme 4         1.448214  0.640379\n                                    Scheme 2         1.987500  0.616942\n                                    Scheme 3         2.703571  0.587277\n                                    Scheme 1         3.860714  0.511964\n                  Lee2019_MI        Scheme 4         1.351786  0.642455\n                                    Scheme 2         1.871429  0.612254\n                                    Scheme 3         2.805357  0.565871\n                                    Scheme 1         3.971429  0.463192\n                  PhysionetMI       Scheme 4         1.430357  0.649687\n                                    Scheme 2         1.583929  0.642098\n                                    Scheme 3         3.042857  0.541295\n                                    Scheme 1         3.942857  0.472009",
      "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></th>\n      <th></th>\n      <th>rank</th>\n      <th>test_acc</th>\n    </tr>\n    <tr>\n      <th>finetune_dataset</th>\n      <th>pretrain_dataset</th>\n      <th>finetune_scheme</th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th rowspan=\"16\" valign=\"top\">BNCI2014_001</th>\n      <th rowspan=\"4\" valign=\"top\">BNCI2014_004</th>\n      <th>Scheme 4</th>\n      <td>1.469444</td>\n      <td>0.496123</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>2.041667</td>\n      <td>0.470120</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.691667</td>\n      <td>0.448457</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.797222</td>\n      <td>0.395197</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">Lee2019_MI</th>\n      <th>Scheme 4</th>\n      <td>1.602778</td>\n      <td>0.495390</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>2.138889</td>\n      <td>0.479398</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.513889</td>\n      <td>0.466667</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.744444</td>\n      <td>0.402836</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">PhysionetMI</th>\n      <th>Scheme 4</th>\n      <td>1.816667</td>\n      <td>0.525463</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>2.072222</td>\n      <td>0.519522</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.502778</td>\n      <td>0.507523</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.608333</td>\n      <td>0.458063</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">Schirrmeister2017</th>\n      <th>Scheme 4</th>\n      <td>1.688889</td>\n      <td>0.448052</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>1.886111</td>\n      <td>0.444541</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.886111</td>\n      <td>0.399711</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.538889</td>\n      <td>0.376485</td>\n    </tr>\n    <tr>\n      <th rowspan=\"16\" valign=\"top\">BNCI2014_004</th>\n      <th rowspan=\"4\" valign=\"top\">BNCI2014_001</th>\n      <th>Scheme 4</th>\n      <td>1.638889</td>\n      <td>0.763973</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>2.013889</td>\n      <td>0.752059</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.658333</td>\n      <td>0.725010</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.688889</td>\n      <td>0.674928</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">Lee2019_MI</th>\n      <th>Scheme 4</th>\n      <td>1.702778</td>\n      <td>0.778242</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>2.066667</td>\n      <td>0.762879</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.641667</td>\n      <td>0.751200</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.588889</td>\n      <td>0.712314</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">PhysionetMI</th>\n      <th>Scheme 4</th>\n      <td>1.586111</td>\n      <td>0.764908</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>1.927778</td>\n      <td>0.754906</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.836111</td>\n      <td>0.726131</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.650000</td>\n      <td>0.695937</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">Schirrmeister2017</th>\n      <th>Scheme 4</th>\n      <td>1.594444</td>\n      <td>0.768765</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>2.011111</td>\n      <td>0.754866</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.633333</td>\n      <td>0.725719</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.761111</td>\n      <td>0.680340</td>\n    </tr>\n    <tr>\n      <th rowspan=\"16\" valign=\"top\">Schirrmeister2017</th>\n      <th rowspan=\"4\" valign=\"top\">BNCI2014_001</th>\n      <th>Scheme 4</th>\n      <td>1.367857</td>\n      <td>0.644531</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>1.841071</td>\n      <td>0.622076</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.860714</td>\n      <td>0.571027</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.930357</td>\n      <td>0.468929</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">BNCI2014_004</th>\n      <th>Scheme 4</th>\n      <td>1.448214</td>\n      <td>0.640379</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>1.987500</td>\n      <td>0.616942</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.703571</td>\n      <td>0.587277</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.860714</td>\n      <td>0.511964</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">Lee2019_MI</th>\n      <th>Scheme 4</th>\n      <td>1.351786</td>\n      <td>0.642455</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>1.871429</td>\n      <td>0.612254</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>2.805357</td>\n      <td>0.565871</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.971429</td>\n      <td>0.463192</td>\n    </tr>\n    <tr>\n      <th rowspan=\"4\" valign=\"top\">PhysionetMI</th>\n      <th>Scheme 4</th>\n      <td>1.430357</td>\n      <td>0.649687</td>\n    </tr>\n    <tr>\n      <th>Scheme 2</th>\n      <td>1.583929</td>\n      <td>0.642098</td>\n    </tr>\n    <tr>\n      <th>Scheme 3</th>\n      <td>3.042857</td>\n      <td>0.541295</td>\n    </tr>\n    <tr>\n      <th>Scheme 1</th>\n      <td>3.942857</td>\n      <td>0.472009</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Rank finetune schemes\n",
    "cols_rank_scheme = [c for c in unique_cols if c != 'finetune_scheme']\n",
    "d = results.copy()\n",
    "d = d[d.pretrain_dataset.ne(d.finetune_dataset)]\n",
    "d['rank'] = d.groupby(cols_rank_scheme, dropna=False).test_acc.rank(ascending=False)\n",
    "d.groupby(['finetune_dataset', 'pretrain_dataset', 'finetune_scheme']).agg(\n",
    "    {'rank': 'mean', 'test_acc': 'mean'}).sort_values(\n",
    "    'rank').sort_index(level=[0, 1], sort_remaining=False)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:37.467635Z",
     "start_time": "2024-02-13T17:08:37.430574Z"
    }
   },
   "id": "136d3462415a66e7",
   "execution_count": 18
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Comparison between commits"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "f25eca5aa945f20a"
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Keeping groups with 3 elements\n"
     ]
    },
    {
     "data": {
      "text/plain": "                              rank  test_acc\ndirectory    ref_channel                    \nresults_v0-1 average      1.855636  0.620554\n             NaN          2.183385  0.602829\nresults_v2   NaN          1.960979  0.612283",
      "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></th>\n      <th>rank</th>\n      <th>test_acc</th>\n    </tr>\n    <tr>\n      <th>directory</th>\n      <th>ref_channel</th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th rowspan=\"2\" valign=\"top\">results_v0-1</th>\n      <th>average</th>\n      <td>1.855636</td>\n      <td>0.620554</td>\n    </tr>\n    <tr>\n      <th>NaN</th>\n      <td>2.183385</td>\n      <td>0.602829</td>\n    </tr>\n    <tr>\n      <th>results_v2</th>\n      <th>NaN</th>\n      <td>1.960979</td>\n      <td>0.612283</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Compare conditions by grouping by pretrain_dataset, finetune_dataset, finetune_subject, finetune_fold, finetune_scheme \n",
    "# Then only keep groups with results from all conditions\n",
    "# then rank test_acc within each group\n",
    "# then compute the mean rank for each condition\n",
    "\n",
    "groups = ['pretrain_dataset', 'finetune_dataset', 'finetune_scheme', 'finetune_subject', 'finetune_fold',\n",
    "          ]\n",
    "d = results_all[results_all.pretrain_dataset.isin(['BNCI2014_004', 'BNCI2014_001', 'Schirrmeister2017'])].groupby(\n",
    "    groups + ['directory', 'ref_channel'], dropna=False).test_acc.mean().reset_index()\n",
    "gs = d.groupby(groups).transform('size')\n",
    "g_count = gs.max()\n",
    "# g_count = 3\n",
    "print(f'Keeping groups with {g_count} elements')\n",
    "d = d[gs == g_count]\n",
    "d['rank'] = d.groupby(groups).test_acc.rank(ascending=False)\n",
    "d.groupby(['finetune_dataset', 'directory', 'ref_channel'], dropna=False).agg(\n",
    "    {'rank': 'mean', 'test_acc': 'mean'}).groupby(\n",
    "    ['directory', 'ref_channel'], dropna=False).agg(\n",
    "    {'rank': 'mean', 'test_acc': 'mean'})\n",
    "\n"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:39.875467Z",
     "start_time": "2024-02-13T17:08:39.825509Z"
    }
   },
   "id": "b253090fbd2337a8",
   "execution_count": 19
  },
  {
   "cell_type": "markdown",
   "source": [],
   "metadata": {
    "collapsed": false
   },
   "id": "213aa7e47ecb9807"
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "directory     ref_channel\nresults_v0-1  average         3200\n              NaN            12800\nresults_v2    NaN             1920\nName: test_acc, dtype: int64"
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results_all.groupby(['directory', 'ref_channel'], dropna=False).test_acc.count()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-13T17:08:41.291385Z",
     "start_time": "2024-02-13T17:08:41.278072Z"
    }
   },
   "id": "b7fe1defc9b7577d",
   "execution_count": 20
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Plots\n",
    "### ranking\n"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "bda2674cbab41203"
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "sns.set_theme(style=\"whitegrid\", context=\"paper\", font_scale=1)"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-15T18:20:52.343446Z",
     "start_time": "2024-02-15T18:20:52.341948Z"
    }
   },
   "id": "4d7a7825b630e6c7",
   "execution_count": 112
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 480x200 with 3 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "finetune_dataset   rank\n",
      "BNCI2014_001       1       180.0\n",
      "                   2       180.0\n",
      "                   3       180.0\n",
      "BNCI2014_004       1       180.0\n",
      "                   2       180.0\n",
      "                   3       180.0\n",
      "Schirrmeister2017  1       280.0\n",
      "                   2       280.0\n",
      "                   3       280.0\n",
      "Name: rank_weight, dtype: float64\n"
     ]
    },
    {
     "data": {
      "text/plain": "finetune_dataset   pretrain_dataset   rank\nBNCI2014_001       BNCI2014_004       1        45.000000\n                                      2        89.500000\n                                      3        45.500000\n                   PhysionetMI        1       115.000000\n                                      2        49.500000\n                                      3        15.500000\n                   Schirrmeister2017  1        20.000000\n                                      2        41.000000\n                                      3       119.000000\nBNCI2014_004       BNCI2014_001       1        43.000000\n                                      2        61.000000\n                                      3        76.000000\n                   Lee2019_MI         1        83.500000\n                                      2        61.500000\n                                      3        35.000000\n                   Schirrmeister2017  1        53.500000\n                                      2        57.500000\n                                      3        69.000000\nSchirrmeister2017  BNCI2014_001       1        91.333333\n                                      2        88.833333\n                                      3        99.833333\n                   BNCI2014_004       1        74.333333\n                                      2       104.833333\n                                      3       100.833333\n                   PhysionetMI        1       114.333333\n                                      2        86.333333\n                                      3        79.333333\nName: rank_weight, dtype: float64"
     },
     "execution_count": 113,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# rank of pretrain datasets\n",
    "df = results.copy()\n",
    "df = df[df.pretrain_dataset.ne(df.finetune_dataset)]\n",
    "df = df[df.finetune_scheme == 'Scheme 4']\n",
    "# not our recommandations:\n",
    "df = df[(df.finetune_dataset != 'BNCI2014_004') | (df.pretrain_dataset != 'PhysionetMI')]\n",
    "df = df[(df.finetune_dataset != 'BNCI2014_001') | (df.pretrain_dataset != 'Lee2019_MI')]\n",
    "df = df[(df.finetune_dataset != 'Schirrmeister2017') | (df.pretrain_dataset != 'Lee2019_MI')]\n",
    "\n",
    "gp = df.groupby(cols_rank_pretrain_dataset, dropna=False).test_acc\n",
    "df['rank_min'] = gp.rank(method='min', ascending=False)\n",
    "df['rank_max'] = gp.rank(method='max', ascending=False)\n",
    "df['rank_weight'] = 1\n",
    "df['rank'] = 0\n",
    "rows = []\n",
    "for _, r in df.iterrows():\n",
    "    rmin = int(r['rank_min'])\n",
    "    rmax = int(r['rank_max'])\n",
    "    for j in range(rmin, rmax + 1):\n",
    "        r1 = r.copy()\n",
    "        r1['rank_weight'] = 1 / (rmax - rmin + 1)\n",
    "        r1['rank'] = int(j)\n",
    "        rows.append(r1)\n",
    "df = pd.DataFrame(rows)\n",
    "n_ranks = int(df['rank'].max())\n",
    "with sns.axes_style('ticks'):\n",
    "    g = sns.displot(\n",
    "        kind='hist', data=df, x=\"rank\", hue=\"pretrain_dataset\", palette='Set2',\n",
    "        multiple=\"stack\",  #stat='probability',\n",
    "        element='step',  # 'poly',\n",
    "        weights='rank_weight',\n",
    "        # discrete=True,\n",
    "        binwidth=1,\n",
    "        binrange=(0.5, n_ranks + 0.5),\n",
    "        col='finetune_dataset',\n",
    "        aspect=.8,\n",
    "        height=2,\n",
    "        edgecolor='w',\n",
    "        facet_kws=dict(legend_out=False),\n",
    "    )\n",
    "sns.move_legend(\n",
    "    g,\n",
    "    title='Pre-training (donor) dataset',\n",
    "    loc=\"lower center\",\n",
    "    bbox_to_anchor=(.5, 1),\n",
    "    ncols=3,\n",
    "    frameon=True,\n",
    ")\n",
    "ticks = range(1, n_ranks + 1) if n_ranks < 10 else range(1, n_ranks + 1, 2)\n",
    "g.set(xticks=ticks)\n",
    "# plt.suptitle('Rank distribution of the pre-training datasets')\n",
    "g.tick_params(bottom=False, top=False)\n",
    "g.set_xlabels('Rank')\n",
    "g.set_ylabels(f'Fold count')\n",
    "g.set_titles(template=\"Test dataset\\n{col_name}\")\n",
    "# g.legend.draw_frame(True)\n",
    "g.savefig('export/rank_pretrain_datasets_scheme4.pdf', bbox_inches='tight')\n",
    "plt.show()\n",
    "print(df.groupby(['finetune_dataset', 'rank']).rank_weight.sum())\n",
    "df.groupby(['finetune_dataset', 'pretrain_dataset', 'rank']).rank_weight.sum()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-15T18:20:53.641048Z",
     "start_time": "2024-02-15T18:20:53.017197Z"
    }
   },
   "id": "c905408afcf252d1",
   "execution_count": 113
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 480x1000 with 15 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# rank of finetune schemes\n",
    "df = results.copy()\n",
    "df = df[df.pretrain_dataset.ne(df.finetune_dataset)]\n",
    "scheme_order = ['Scheme 4', 'Scheme 2', 'Scheme 3', 'Scheme 1']\n",
    "# df.finetune_scheme = df.finetune_scheme.astype(pd.CategoricalDtype(categories=scheme_order))\n",
    "\n",
    "gp = df.groupby(cols_rank_scheme, dropna=False).test_acc\n",
    "df['rank_min'] = gp.rank(method='min', ascending=False)\n",
    "df['rank_max'] = gp.rank(method='max', ascending=False)\n",
    "df['rank_weight'] = 1\n",
    "df['rank'] = 0\n",
    "rows = []\n",
    "for _, r in df.iterrows():\n",
    "    rmin = int(r['rank_min'])\n",
    "    rmax = int(r['rank_max'])\n",
    "    for j in range(rmin, rmax + 1):\n",
    "        r1 = r.copy()\n",
    "        r1['rank_weight'] = 1 / (rmax - rmin + 1)\n",
    "        r1['rank'] = int(j)\n",
    "        rows.append(r1)\n",
    "df = pd.DataFrame(rows)\n",
    "n_ranks = int(df['rank'].max())\n",
    "with sns.axes_style('ticks'):\n",
    "    g = sns.displot(\n",
    "        kind='hist', data=df, x=\"rank\", hue=\"finetune_scheme\", palette='Set2',\n",
    "        multiple=\"stack\",  #stat='probability',\n",
    "        element='step',  # 'poly',\n",
    "        # discrete=True,\n",
    "        weights='rank_weight',\n",
    "        binwidth=1,\n",
    "        binrange=(0.5, n_ranks + 0.5),\n",
    "        col='finetune_dataset',\n",
    "        row='pretrain_dataset',\n",
    "        hue_order=scheme_order,\n",
    "        aspect=.8,\n",
    "        height=2,\n",
    "        edgecolor='w',\n",
    "        facet_kws=dict(legend_out=False),\n",
    "    )\n",
    "sns.move_legend(\n",
    "    g,\n",
    "    title='Fine-tuning scheme',\n",
    "    loc=\"lower center\",\n",
    "    bbox_to_anchor=(.5, 1),\n",
    "    ncols=4,\n",
    "    frameon=True,\n",
    ")\n",
    "plt.xticks(range(1, n_ranks + 1) if n_ranks < 10 else range(1, n_ranks + 1, 2))\n",
    "# plt.suptitle('Rank distribution of the pre-training datasets')\n",
    "g.set_titles(template=\"From {row_name}\\n to {col_name}\")\n",
    "g.set_xlabels('Rank')\n",
    "g.set_ylabels(f'Fold count')\n",
    "g.savefig('export/rank_finetune_schemes.pdf', bbox_inches='tight')\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-15T18:21:05.088761Z",
     "start_time": "2024-02-15T18:21:02.253780Z"
    }
   },
   "id": "7939fedcaaa8a712",
   "execution_count": 114
  },
  {
   "cell_type": "markdown",
   "source": [
    "### Barplots"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "b398f08fe73f527"
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 814.278x200 with 3 Axes>",
      "image/png": "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"
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df = results.copy()\n",
    "df = df[df.pretrain_dataset.ne(df.finetune_dataset)]\n",
    "df = df[df.finetune_scheme == 'Scheme 4']\n",
    "# not our recommandations:\n",
    "df = df[(df.finetune_dataset != 'BNCI2014_004') | (df.pretrain_dataset != 'PhysionetMI')]\n",
    "df = df[(df.finetune_dataset != 'BNCI2014_001') | (df.pretrain_dataset != 'Lee2019_MI')]\n",
    "df = df[(df.finetune_dataset != 'Schirrmeister2017') | (df.pretrain_dataset != 'Lee2019_MI')]\n",
    "legend_label = 'Pre-training (donor) dataset colour'\n",
    "df[legend_label] = 'Used by Xie et al. (2023)'\n",
    "df.loc[df.pretrain_dataset.isin(['PhysionetMI', 'Lee2019_MI']), legend_label] = 'Our recommendation'\n",
    "\n",
    "g = sns.catplot(\n",
    "    data=df, x='test_acc', y='pretrain_dataset', hue=legend_label, kind='bar', errorbar='sd',\n",
    "    col='finetune_dataset', dodge=False,\n",
    "    aspect=1,\n",
    "    height=2,\n",
    "    palette='Paired',\n",
    "    facet_kws=dict(legend_out=False),\n",
    ")\n",
    "sns.move_legend(\n",
    "    g,\n",
    "    loc=\"lower center\",\n",
    "    bbox_to_anchor=(.5, 1),\n",
    "    ncols=2,\n",
    "    frameon=True,\n",
    ")\n",
    "ticks = [0, 0.2, 0.4, 0.6, 0.8]\n",
    "g.set(xticks=ticks)\n",
    "g.set_titles(col_template=\"Test dataset\\n{col_name}\")\n",
    "g.set_xlabels('Test accuracy')\n",
    "g.set_ylabels('Pre-training dataset')\n",
    "g.savefig('export/barplot_scheme4.pdf', bbox_inches='tight')\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-15T18:21:05.564400Z",
     "start_time": "2024-02-15T18:21:05.091992Z"
    }
   },
   "id": "905269a865fd240b",
   "execution_count": 115
  },
  {
   "cell_type": "markdown",
   "source": [
    "## Statistical tests"
   ],
   "metadata": {
    "collapsed": false
   },
   "id": "ad55e7246e3bb2da"
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "finetune_dataset                                        Schirrmeister2017  \\\npretrain_dataset                                             BNCI2014_001   \nfinetune_scheme                                                  Scheme 1   \nfinetune_subject finetune_fold directory    ref_channel                     \n1                0             results_v0-1 NaN                   0.42500   \n                 1             results_v0-1 NaN                   0.45000   \n                 2             results_v0-1 NaN                   0.46875   \n                 3             results_v0-1 NaN                   0.46875   \n                 4             results_v0-1 NaN                   0.48125   \n...                                                                   ...   \n14               15            results_v0-1 NaN                   0.49375   \n                 16            results_v0-1 NaN                   0.45625   \n                 17            results_v0-1 NaN                   0.46875   \n                 18            results_v0-1 NaN                   0.49375   \n                 19            results_v0-1 NaN                   0.53750   \n\nfinetune_dataset                                                           \\\npretrain_dataset                                                            \nfinetune_scheme                                         Scheme 2 Scheme 3   \nfinetune_subject finetune_fold directory    ref_channel                     \n1                0             results_v0-1 NaN          0.47500  0.47500   \n                 1             results_v0-1 NaN          0.55625  0.49375   \n                 2             results_v0-1 NaN          0.60625  0.51250   \n                 3             results_v0-1 NaN          0.56250  0.51250   \n                 4             results_v0-1 NaN          0.55000  0.50000   \n...                                                          ...      ...   \n14               15            results_v0-1 NaN          0.68750  0.51250   \n                 16            results_v0-1 NaN          0.72500  0.55000   \n                 17            results_v0-1 NaN          0.71250  0.56875   \n                 18            results_v0-1 NaN          0.68750  0.54375   \n                 19            results_v0-1 NaN          0.67500  0.54375   \n\nfinetune_dataset                                                 BNCI2014_004  \\\npretrain_dataset                                                 BNCI2014_001   \nfinetune_scheme                                         Scheme 4     Scheme 1   \nfinetune_subject finetune_fold directory    ref_channel                         \n1                0             results_v0-1 NaN          0.52500     0.615625   \n                 1             results_v0-1 NaN          0.60625     0.562500   \n                 2             results_v0-1 NaN          0.60000     0.584375   \n                 3             results_v0-1 NaN          0.56250     0.609375   \n                 4             results_v0-1 NaN          0.53125     0.590625   \n...                                                          ...          ...   \n14               15            results_v0-1 NaN          0.70625          NaN   \n                 16            results_v0-1 NaN          0.66250          NaN   \n                 17            results_v0-1 NaN          0.80000          NaN   \n                 18            results_v0-1 NaN          0.74375          NaN   \n                 19            results_v0-1 NaN          0.73125          NaN   \n\nfinetune_dataset                                                             \\\npretrain_dataset                                                              \nfinetune_scheme                                          Scheme 2  Scheme 3   \nfinetune_subject finetune_fold directory    ref_channel                       \n1                0             results_v0-1 NaN          0.718750  0.634375   \n                 1             results_v0-1 NaN          0.718750  0.568750   \n                 2             results_v0-1 NaN          0.715625  0.662500   \n                 3             results_v0-1 NaN          0.706250  0.659375   \n                 4             results_v0-1 NaN          0.709375  0.687500   \n...                                                           ...       ...   \n14               15            results_v0-1 NaN               NaN       NaN   \n                 16            results_v0-1 NaN               NaN       NaN   \n                 17            results_v0-1 NaN               NaN       NaN   \n                 18            results_v0-1 NaN               NaN       NaN   \n                 19            results_v0-1 NaN               NaN       NaN   \n\nfinetune_dataset                                                   \\\npretrain_dataset                                                    \nfinetune_scheme                                          Scheme 4   \nfinetune_subject finetune_fold directory    ref_channel             \n1                0             results_v0-1 NaN          0.725000   \n                 1             results_v0-1 NaN          0.584375   \n                 2             results_v0-1 NaN          0.725000   \n                 3             results_v0-1 NaN          0.706250   \n                 4             results_v0-1 NaN          0.728125   \n...                                                           ...   \n14               15            results_v0-1 NaN               NaN   \n                 16            results_v0-1 NaN               NaN   \n                 17            results_v0-1 NaN               NaN   \n                 18            results_v0-1 NaN               NaN   \n                 19            results_v0-1 NaN               NaN   \n\nfinetune_dataset                                        BNCI2014_001  \\\npretrain_dataset                                        BNCI2014_004   \nfinetune_scheme                                             Scheme 1   \nfinetune_subject finetune_fold directory    ref_channel                \n1                0             results_v0-1 NaN             0.500000   \n                 1             results_v0-1 NaN             0.493056   \n                 2             results_v0-1 NaN             0.500000   \n                 3             results_v0-1 NaN             0.534722   \n                 4             results_v0-1 NaN             0.517361   \n...                                                              ...   \n14               15            results_v0-1 NaN                  NaN   \n                 16            results_v0-1 NaN                  NaN   \n                 17            results_v0-1 NaN                  NaN   \n                 18            results_v0-1 NaN                  NaN   \n                 19            results_v0-1 NaN                  NaN   \n\nfinetune_dataset                                                   ...  \\\npretrain_dataset                                                   ...   \nfinetune_scheme                                          Scheme 2  ...   \nfinetune_subject finetune_fold directory    ref_channel            ...   \n1                0             results_v0-1 NaN          0.631944  ...   \n                 1             results_v0-1 NaN          0.649306  ...   \n                 2             results_v0-1 NaN          0.642361  ...   \n                 3             results_v0-1 NaN          0.645833  ...   \n                 4             results_v0-1 NaN          0.638889  ...   \n...                                                           ...  ...   \n14               15            results_v0-1 NaN               NaN  ...   \n                 16            results_v0-1 NaN               NaN  ...   \n                 17            results_v0-1 NaN               NaN  ...   \n                 18            results_v0-1 NaN               NaN  ...   \n                 19            results_v0-1 NaN               NaN  ...   \n\nfinetune_dataset                                        BNCI2014_004  \\\npretrain_dataset                                          Lee2019_MI   \nfinetune_scheme                                             Scheme 3   \nfinetune_subject finetune_fold directory    ref_channel                \n1                0             results_v0-1 NaN             0.659375   \n                 1             results_v0-1 NaN             0.684375   \n                 2             results_v0-1 NaN             0.684375   \n                 3             results_v0-1 NaN             0.703125   \n                 4             results_v0-1 NaN             0.700000   \n...                                                              ...   \n14               15            results_v0-1 NaN                  NaN   \n                 16            results_v0-1 NaN                  NaN   \n                 17            results_v0-1 NaN                  NaN   \n                 18            results_v0-1 NaN                  NaN   \n                 19            results_v0-1 NaN                  NaN   \n\nfinetune_dataset                                                   \\\npretrain_dataset                                                    \nfinetune_scheme                                          Scheme 4   \nfinetune_subject finetune_fold directory    ref_channel             \n1                0             results_v0-1 NaN          0.737500   \n                 1             results_v0-1 NaN          0.746875   \n                 2             results_v0-1 NaN          0.687500   \n                 3             results_v0-1 NaN          0.737500   \n                 4             results_v0-1 NaN          0.775000   \n...                                                           ...   \n14               15            results_v0-1 NaN               NaN   \n                 16            results_v0-1 NaN               NaN   \n                 17            results_v0-1 NaN               NaN   \n                 18            results_v0-1 NaN               NaN   \n                 19            results_v0-1 NaN               NaN   \n\nfinetune_dataset                                             BNCI2014_001  \\\npretrain_dataset                                        Schirrmeister2017   \nfinetune_scheme                                                  Scheme 1   \nfinetune_subject finetune_fold directory    ref_channel                     \n1                0             results_v0-1 NaN                  0.378472   \n                 1             results_v0-1 NaN                  0.322917   \n                 2             results_v0-1 NaN                  0.368056   \n                 3             results_v0-1 NaN                  0.368056   \n                 4             results_v0-1 NaN                  0.371528   \n...                                                                   ...   \n14               15            results_v0-1 NaN                       NaN   \n                 16            results_v0-1 NaN                       NaN   \n                 17            results_v0-1 NaN                       NaN   \n                 18            results_v0-1 NaN                       NaN   \n                 19            results_v0-1 NaN                       NaN   \n\nfinetune_dataset                                                             \\\npretrain_dataset                                                              \nfinetune_scheme                                          Scheme 2  Scheme 3   \nfinetune_subject finetune_fold directory    ref_channel                       \n1                0             results_v0-1 NaN          0.531250  0.423611   \n                 1             results_v0-1 NaN          0.510417  0.430556   \n                 2             results_v0-1 NaN          0.597222  0.465278   \n                 3             results_v0-1 NaN          0.451389  0.420139   \n                 4             results_v0-1 NaN          0.583333  0.440972   \n...                                                           ...       ...   \n14               15            results_v0-1 NaN               NaN       NaN   \n                 16            results_v0-1 NaN               NaN       NaN   \n                 17            results_v0-1 NaN               NaN       NaN   \n                 18            results_v0-1 NaN               NaN       NaN   \n                 19            results_v0-1 NaN               NaN       NaN   \n\nfinetune_dataset                                                   \\\npretrain_dataset                                                    \nfinetune_scheme                                          Scheme 4   \nfinetune_subject finetune_fold directory    ref_channel             \n1                0             results_v0-1 NaN          0.486111   \n                 1             results_v0-1 NaN          0.534722   \n                 2             results_v0-1 NaN          0.597222   \n                 3             results_v0-1 NaN          0.482639   \n                 4             results_v0-1 NaN          0.572917   \n...                                                           ...   \n14               15            results_v0-1 NaN               NaN   \n                 16            results_v0-1 NaN               NaN   \n                 17            results_v0-1 NaN               NaN   \n                 18            results_v0-1 NaN               NaN   \n                 19            results_v0-1 NaN               NaN   \n\nfinetune_dataset                                        Schirrmeister2017  \\\npretrain_dataset                                              PhysionetMI   \nfinetune_scheme                                                  Scheme 1   \nfinetune_subject finetune_fold directory    ref_channel                     \n1                0             results_v0-1 NaN                   0.44375   \n                 1             results_v0-1 NaN                   0.39375   \n                 2             results_v0-1 NaN                   0.38750   \n                 3             results_v0-1 NaN                   0.40000   \n                 4             results_v0-1 NaN                   0.43125   \n...                                                                   ...   \n14               15            results_v0-1 NaN                   0.42500   \n                 16            results_v0-1 NaN                   0.45625   \n                 17            results_v0-1 NaN                   0.42500   \n                 18            results_v0-1 NaN                   0.45000   \n                 19            results_v0-1 NaN                   0.42500   \n\nfinetune_dataset                                                           \\\npretrain_dataset                                                            \nfinetune_scheme                                         Scheme 2 Scheme 3   \nfinetune_subject finetune_fold directory    ref_channel                     \n1                0             results_v0-1 NaN          0.60625  0.50000   \n                 1             results_v0-1 NaN          0.52500  0.48125   \n                 2             results_v0-1 NaN          0.46875  0.47500   \n                 3             results_v0-1 NaN          0.60000  0.47500   \n                 4             results_v0-1 NaN          0.53750  0.50000   \n...                                                          ...      ...   \n14               15            results_v0-1 NaN          0.66250  0.48125   \n                 16            results_v0-1 NaN          0.63125  0.50625   \n                 17            results_v0-1 NaN          0.64375  0.48750   \n                 18            results_v0-1 NaN          0.66250  0.50000   \n                 19            results_v0-1 NaN          0.63750  0.50000   \n\nfinetune_dataset                                                  \npretrain_dataset                                                  \nfinetune_scheme                                         Scheme 4  \nfinetune_subject finetune_fold directory    ref_channel           \n1                0             results_v0-1 NaN          0.59375  \n                 1             results_v0-1 NaN          0.53125  \n                 2             results_v0-1 NaN          0.52500  \n                 3             results_v0-1 NaN          0.62500  \n                 4             results_v0-1 NaN          0.52500  \n...                                                          ...  \n14               15            results_v0-1 NaN          0.70000  \n                 16            results_v0-1 NaN          0.66875  \n                 17            results_v0-1 NaN          0.68125  \n                 18            results_v0-1 NaN          0.68125  \n                 19            results_v0-1 NaN          0.71875  \n\n[280 rows x 60 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 tr th {\n        text-align: left;\n    }\n\n    .dataframe thead tr:last-of-type th {\n        text-align: right;\n    }\n</style>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th>finetune_dataset</th>\n      <th colspan=\"4\" halign=\"left\">Schirrmeister2017</th>\n      <th colspan=\"4\" halign=\"left\">BNCI2014_004</th>\n      <th colspan=\"2\" halign=\"left\">BNCI2014_001</th>\n      <th>...</th>\n      <th colspan=\"2\" halign=\"left\">BNCI2014_004</th>\n      <th colspan=\"4\" halign=\"left\">BNCI2014_001</th>\n      <th colspan=\"4\" halign=\"left\">Schirrmeister2017</th>\n    </tr>\n    <tr>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th>pretrain_dataset</th>\n      <th colspan=\"4\" halign=\"left\">BNCI2014_001</th>\n      <th colspan=\"4\" halign=\"left\">BNCI2014_001</th>\n      <th colspan=\"2\" halign=\"left\">BNCI2014_004</th>\n      <th>...</th>\n      <th colspan=\"2\" halign=\"left\">Lee2019_MI</th>\n      <th colspan=\"4\" halign=\"left\">Schirrmeister2017</th>\n      <th colspan=\"4\" halign=\"left\">PhysionetMI</th>\n    </tr>\n    <tr>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th>finetune_scheme</th>\n      <th>Scheme 1</th>\n      <th>Scheme 2</th>\n      <th>Scheme 3</th>\n      <th>Scheme 4</th>\n      <th>Scheme 1</th>\n      <th>Scheme 2</th>\n      <th>Scheme 3</th>\n      <th>Scheme 4</th>\n      <th>Scheme 1</th>\n      <th>Scheme 2</th>\n      <th>...</th>\n      <th>Scheme 3</th>\n      <th>Scheme 4</th>\n      <th>Scheme 1</th>\n      <th>Scheme 2</th>\n      <th>Scheme 3</th>\n      <th>Scheme 4</th>\n      <th>Scheme 1</th>\n      <th>Scheme 2</th>\n      <th>Scheme 3</th>\n      <th>Scheme 4</th>\n    </tr>\n    <tr>\n      <th>finetune_subject</th>\n      <th>finetune_fold</th>\n      <th>directory</th>\n      <th>ref_channel</th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n      <th></th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th rowspan=\"5\" valign=\"top\">1</th>\n      <th>0</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.42500</td>\n      <td>0.47500</td>\n      <td>0.47500</td>\n      <td>0.52500</td>\n      <td>0.615625</td>\n      <td>0.718750</td>\n      <td>0.634375</td>\n      <td>0.725000</td>\n      <td>0.500000</td>\n      <td>0.631944</td>\n      <td>...</td>\n      <td>0.659375</td>\n      <td>0.737500</td>\n      <td>0.378472</td>\n      <td>0.531250</td>\n      <td>0.423611</td>\n      <td>0.486111</td>\n      <td>0.44375</td>\n      <td>0.60625</td>\n      <td>0.50000</td>\n      <td>0.59375</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.45000</td>\n      <td>0.55625</td>\n      <td>0.49375</td>\n      <td>0.60625</td>\n      <td>0.562500</td>\n      <td>0.718750</td>\n      <td>0.568750</td>\n      <td>0.584375</td>\n      <td>0.493056</td>\n      <td>0.649306</td>\n      <td>...</td>\n      <td>0.684375</td>\n      <td>0.746875</td>\n      <td>0.322917</td>\n      <td>0.510417</td>\n      <td>0.430556</td>\n      <td>0.534722</td>\n      <td>0.39375</td>\n      <td>0.52500</td>\n      <td>0.48125</td>\n      <td>0.53125</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.46875</td>\n      <td>0.60625</td>\n      <td>0.51250</td>\n      <td>0.60000</td>\n      <td>0.584375</td>\n      <td>0.715625</td>\n      <td>0.662500</td>\n      <td>0.725000</td>\n      <td>0.500000</td>\n      <td>0.642361</td>\n      <td>...</td>\n      <td>0.684375</td>\n      <td>0.687500</td>\n      <td>0.368056</td>\n      <td>0.597222</td>\n      <td>0.465278</td>\n      <td>0.597222</td>\n      <td>0.38750</td>\n      <td>0.46875</td>\n      <td>0.47500</td>\n      <td>0.52500</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.46875</td>\n      <td>0.56250</td>\n      <td>0.51250</td>\n      <td>0.56250</td>\n      <td>0.609375</td>\n      <td>0.706250</td>\n      <td>0.659375</td>\n      <td>0.706250</td>\n      <td>0.534722</td>\n      <td>0.645833</td>\n      <td>...</td>\n      <td>0.703125</td>\n      <td>0.737500</td>\n      <td>0.368056</td>\n      <td>0.451389</td>\n      <td>0.420139</td>\n      <td>0.482639</td>\n      <td>0.40000</td>\n      <td>0.60000</td>\n      <td>0.47500</td>\n      <td>0.62500</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.48125</td>\n      <td>0.55000</td>\n      <td>0.50000</td>\n      <td>0.53125</td>\n      <td>0.590625</td>\n      <td>0.709375</td>\n      <td>0.687500</td>\n      <td>0.728125</td>\n      <td>0.517361</td>\n      <td>0.638889</td>\n      <td>...</td>\n      <td>0.700000</td>\n      <td>0.775000</td>\n      <td>0.371528</td>\n      <td>0.583333</td>\n      <td>0.440972</td>\n      <td>0.572917</td>\n      <td>0.43125</td>\n      <td>0.53750</td>\n      <td>0.50000</td>\n      <td>0.52500</td>\n    </tr>\n    <tr>\n      <th>...</th>\n      <th>...</th>\n      <th>...</th>\n      <th>...</th>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n      <td>...</td>\n    </tr>\n    <tr>\n      <th rowspan=\"5\" valign=\"top\">14</th>\n      <th>15</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.49375</td>\n      <td>0.68750</td>\n      <td>0.51250</td>\n      <td>0.70625</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>0.42500</td>\n      <td>0.66250</td>\n      <td>0.48125</td>\n      <td>0.70000</td>\n    </tr>\n    <tr>\n      <th>16</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.45625</td>\n      <td>0.72500</td>\n      <td>0.55000</td>\n      <td>0.66250</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>0.45625</td>\n      <td>0.63125</td>\n      <td>0.50625</td>\n      <td>0.66875</td>\n    </tr>\n    <tr>\n      <th>17</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.46875</td>\n      <td>0.71250</td>\n      <td>0.56875</td>\n      <td>0.80000</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>0.42500</td>\n      <td>0.64375</td>\n      <td>0.48750</td>\n      <td>0.68125</td>\n    </tr>\n    <tr>\n      <th>18</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.49375</td>\n      <td>0.68750</td>\n      <td>0.54375</td>\n      <td>0.74375</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>0.45000</td>\n      <td>0.66250</td>\n      <td>0.50000</td>\n      <td>0.68125</td>\n    </tr>\n    <tr>\n      <th>19</th>\n      <th>results_v0-1</th>\n      <th>NaN</th>\n      <td>0.53750</td>\n      <td>0.67500</td>\n      <td>0.54375</td>\n      <td>0.73125</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>...</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>NaN</td>\n      <td>0.42500</td>\n      <td>0.63750</td>\n      <td>0.50000</td>\n      <td>0.71875</td>\n    </tr>\n  </tbody>\n</table>\n<p>280 rows × 60 columns</p>\n</div>"
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "columns_stats = ['finetune_dataset', 'pretrain_dataset', 'finetune_scheme']\n",
    "index_stats = [c for c in unique_cols if c not in columns_stats]\n",
    "df_stats = results.pivot(index=index_stats, columns=columns_stats, values='test_acc', )\n",
    "df_stats"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-15T18:39:31.095019Z",
     "start_time": "2024-02-15T18:39:31.039609Z"
    }
   },
   "id": "c73267c4ed57a452",
   "execution_count": 116
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "tests_list = [\n",
    "    # (finetune_dataset, their pretrain_dataset, my pretrain_dataset suggestion),\n",
    "    ('BNCI2014_001', 'Schirrmeister2017', 'PhysionetMI'),\n",
    "    ('BNCI2014_001', 'BNCI2014_004', 'PhysionetMI'),\n",
    "    ('BNCI2014_004', 'Schirrmeister2017', 'Lee2019_MI'),\n",
    "    ('BNCI2014_004', 'BNCI2014_001', 'Lee2019_MI'),\n",
    "    ('Schirrmeister2017', 'BNCI2014_001', 'PhysionetMI'),\n",
    "    ('Schirrmeister2017', 'BNCI2014_004', 'PhysionetMI'),\n",
    "]"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-15T18:39:31.525685Z",
     "start_time": "2024-02-15T18:39:31.411296Z"
    }
   },
   "id": "2d29d1b9c8c248d8",
   "execution_count": 117
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [
    "def test_wilcoxon(df, finetune_dataset, their_pretrain_dataset, my_pretrain_dataset):\n",
    "    ''' verifies if my pretraining dataset suggestion is better than the one used in Xie2023 \n",
    "    for a given target finetuning dataset '''\n",
    "    return wilcoxon(\n",
    "        x=df[(finetune_dataset, my_pretrain_dataset, 'Scheme 4')].dropna(),  # My suggestion\n",
    "        y=df[(finetune_dataset, their_pretrain_dataset, 'Scheme 4')].dropna(),  # Used in Xie2023\n",
    "        alternative='greater',  # H: my suggestion (x) is better than the one used in Xie2023 (y)\n",
    "    )"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-15T18:39:32.337421Z",
     "start_time": "2024-02-15T18:39:32.333623Z"
    }
   },
   "id": "e9412667ec82353",
   "execution_count": 118
  },
  {
   "cell_type": "code",
   "outputs": [
    {
     "data": {
      "text/plain": "    finetune_dataset their_pretrain_dataset my_pretrain_dataset        pvalue  \\\n0       BNCI2014_001      Schirrmeister2017         PhysionetMI  7.356550e-24   \n1       BNCI2014_001           BNCI2014_004         PhysionetMI  1.083432e-10   \n2       BNCI2014_004      Schirrmeister2017          Lee2019_MI  1.312820e-04   \n3       BNCI2014_004           BNCI2014_001          Lee2019_MI  2.926137e-08   \n4  Schirrmeister2017           BNCI2014_001         PhysionetMI  2.177847e-02   \n5  Schirrmeister2017           BNCI2014_004         PhysionetMI  1.845486e-03   \n\n   pvalue_ajusted  \n0    4.413930e-23  \n1    3.250297e-10  \n2    1.969231e-04  \n3    5.852274e-08  \n4    2.177847e-02  \n5    2.214584e-03  ",
      "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>finetune_dataset</th>\n      <th>their_pretrain_dataset</th>\n      <th>my_pretrain_dataset</th>\n      <th>pvalue</th>\n      <th>pvalue_ajusted</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>BNCI2014_001</td>\n      <td>Schirrmeister2017</td>\n      <td>PhysionetMI</td>\n      <td>7.356550e-24</td>\n      <td>4.413930e-23</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>BNCI2014_001</td>\n      <td>BNCI2014_004</td>\n      <td>PhysionetMI</td>\n      <td>1.083432e-10</td>\n      <td>3.250297e-10</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>BNCI2014_004</td>\n      <td>Schirrmeister2017</td>\n      <td>Lee2019_MI</td>\n      <td>1.312820e-04</td>\n      <td>1.969231e-04</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>BNCI2014_004</td>\n      <td>BNCI2014_001</td>\n      <td>Lee2019_MI</td>\n      <td>2.926137e-08</td>\n      <td>5.852274e-08</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>Schirrmeister2017</td>\n      <td>BNCI2014_001</td>\n      <td>PhysionetMI</td>\n      <td>2.177847e-02</td>\n      <td>2.177847e-02</td>\n    </tr>\n    <tr>\n      <th>5</th>\n      <td>Schirrmeister2017</td>\n      <td>BNCI2014_004</td>\n      <td>PhysionetMI</td>\n      <td>1.845486e-03</td>\n      <td>2.214584e-03</td>\n    </tr>\n  </tbody>\n</table>\n</div>"
     },
     "execution_count": 119,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_results = [test_wilcoxon(df_stats, *t) for t in tests_list]\n",
    "ps = [r.pvalue for r in test_results]\n",
    "ps_ajusted = false_discovery_control(ps, method='bh')  # Benjamini-Hochberg correction \n",
    "finetune_datasets, their_pretrain_datasets, my_pretrain_datasets = zip(*tests_list)\n",
    "pd.DataFrame({'finetune_dataset': finetune_datasets, 'their_pretrain_dataset': their_pretrain_datasets,\n",
    "              'my_pretrain_dataset': my_pretrain_datasets, 'pvalue': ps, 'pvalue_ajusted': ps_ajusted})\n",
    "\n"
   ],
   "metadata": {
    "collapsed": false,
    "ExecuteTime": {
     "end_time": "2024-02-15T18:39:33.084570Z",
     "start_time": "2024-02-15T18:39:33.061670Z"
    }
   },
   "id": "81292b428da4a26a",
   "execution_count": 119
  },
  {
   "cell_type": "code",
   "outputs": [],
   "source": [],
   "metadata": {
    "collapsed": false
   },
   "id": "9c6718abe95f16c6"
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}