{
  "nbformat_minor": 0, 
  "nbformat": 4, 
  "cells": [
    {
      "execution_count": null, 
      "cell_type": "code", 
      "source": [
        "%matplotlib inline"
      ], 
      "outputs": [], 
      "metadata": {
        "collapsed": false
      }
    }, 
    {
      "source": [
        "\n==============================================\nReal-time feedback for decoding :: Server Side\n==============================================\n\nThis example demonstrates how to setup a real-time feedback\nmechanism using StimServer and StimClient.\n\nThe idea here is to display future stimuli for the class which\nis predicted less accurately. This allows on-demand adaptation\nof the stimuli depending on the needs of the classifier.\n\nTo run this example, open ipython in two separate terminals.\nIn the first, run rt_feedback_server.py and then wait for the\nmessage\n\n    RtServer: Start\n\nOnce that appears, run rt_feedback_client.py in the other terminal\nand the feedback script should start.\n\nAll brain responses are simulated from a fiff file to make it easy\nto test. However, it should be possible to adapt this script\nfor a real experiment.\n\n\n"
      ], 
      "cell_type": "markdown", 
      "metadata": {}
    }, 
    {
      "execution_count": null, 
      "cell_type": "code", 
      "source": [
        "# Author: Mainak Jas <mainak@neuro.hut.fi>\n#\n# License: BSD (3-clause)\n\nimport time\n\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nfrom sklearn import preprocessing\nfrom sklearn.svm import SVC\nfrom sklearn.pipeline import Pipeline\nfrom sklearn.cross_validation import train_test_split\nfrom sklearn.metrics import confusion_matrix\n\nimport mne\nfrom mne.datasets import sample\nfrom mne.realtime import StimServer\nfrom mne.realtime import MockRtClient\nfrom mne.decoding import Vectorizer, FilterEstimator\n\nprint(__doc__)\n\n# Load fiff file to simulate data\ndata_path = sample.data_path()\nraw_fname = data_path + '/MEG/sample/sample_audvis_filt-0-40_raw.fif'\nraw = mne.io.read_raw_fif(raw_fname, preload=True)\n\n# Instantiating stimulation server\n\n# The with statement is necessary to ensure a clean exit\nwith StimServer(port=4218) as stim_server:\n\n    # The channels to be used while decoding\n    picks = mne.pick_types(raw.info, meg='grad', eeg=False, eog=True,\n                           stim=True, exclude=raw.info['bads'])\n\n    rt_client = MockRtClient(raw)\n\n    # Constructing the pipeline for classification\n    filt = FilterEstimator(raw.info, 1, 40)\n    scaler = preprocessing.StandardScaler()\n    vectorizer = Vectorizer()\n    clf = SVC(C=1, kernel='linear')\n\n    concat_classifier = Pipeline([('filter', filt), ('vector', vectorizer),\n                                  ('scaler', scaler), ('svm', clf)])\n\n    stim_server.start(verbose=True)\n\n    # Just some initially decided events to be simulated\n    # Rest will decided on the fly\n    ev_list = [4, 3, 4, 3, 4, 3, 4, 3, 4, 3, 4]\n\n    score_c1, score_c2, score_x = [], [], []\n\n    for ii in range(50):\n\n        # Tell the stim_client about the next stimuli\n        stim_server.add_trigger(ev_list[ii])\n\n        # Collecting data\n        if ii == 0:\n            X = rt_client.get_event_data(event_id=ev_list[ii], tmin=-0.2,\n                                         tmax=0.5, picks=picks,\n                                         stim_channel='STI 014')[None, ...]\n            y = ev_list[ii]\n        else:\n            X_temp = rt_client.get_event_data(event_id=ev_list[ii], tmin=-0.2,\n                                              tmax=0.5, picks=picks,\n                                              stim_channel='STI 014')\n            X_temp = X_temp[np.newaxis, ...]\n\n            X = np.concatenate((X, X_temp), axis=0)\n\n            time.sleep(1)  # simulating the isi\n            y = np.append(y, ev_list[ii])\n\n        # Start decoding after collecting sufficient data\n        if ii >= 10:\n            # Now start doing rtfeedback\n            X_train, X_test, y_train, y_test = train_test_split(X, y,\n                                                                test_size=0.2,\n                                                                random_state=7)\n\n            y_pred = concat_classifier.fit(X_train, y_train).predict(X_test)\n\n            cm = confusion_matrix(y_test, y_pred)\n\n            score_c1.append(float(cm[0, 0]) / sum(cm, 1)[0] * 100)\n            score_c2.append(float(cm[1, 1]) / sum(cm, 1)[1] * 100)\n\n            # do something if one class is decoded better than the other\n            if score_c1[-1] < score_c2[-1]:\n                print(\"We decoded class RV better than class LV\")\n                ev_list.append(3)  # adding more LV to future simulated data\n            else:\n                print(\"We decoded class LV better than class RV\")\n                ev_list.append(4)  # adding more RV to future simulated data\n\n            # Clear the figure\n            plt.clf()\n\n            # The x-axis for the plot\n            score_x.append(ii)\n\n            # Now plot the accuracy\n            plt.plot(score_x[-5:], score_c1[-5:])\n            plt.hold(True)\n            plt.plot(score_x[-5:], score_c2[-5:])\n            plt.xlabel('Trials')\n            plt.ylabel('Classification score (% correct)')\n            plt.title('Real-time feedback')\n            plt.ylim([0, 100])\n            plt.xticks(score_x[-5:])\n            plt.legend(('LV', 'RV'), loc='upper left')\n            plt.show()"
      ], 
      "outputs": [], 
      "metadata": {
        "collapsed": false
      }
    }
  ], 
  "metadata": {
    "kernelspec": {
      "display_name": "Python 2", 
      "name": "python2", 
      "language": "python"
    }, 
    "language_info": {
      "mimetype": "text/x-python", 
      "nbconvert_exporter": "python", 
      "name": "python", 
      "file_extension": ".py", 
      "version": "2.7.13", 
      "pygments_lexer": "ipython2", 
      "codemirror_mode": {
        "version": 2, 
        "name": "ipython"
      }
    }
  }
}