History: Task1: Seperation of dependent components

Comparing version 26 with version 35

Causal Analysis of simulated EEG data


Description of task


Motivation: Noninvasive electrophysiological measurements like EEG/MEG measure to large extent unknown superpositions of very many sources. Any relation observed between channels is dominated by meaningless mixtures of mainly independent sources. The question is how to observe and properly interpret true interactions in the presence of such strong confounders.

Download data (external link) here.

To read the data into MATLAB, type
fid=fopen('simuldata.bin');
data=reshape(fread(fid,'float'),6000,2,1000);

The data consists of 1000 examples of bivariate data for 6000 time points. Each example is a superposition of a signal (of interest) and noise. The signal is constructed from a unidirectional bivariate AR-model of order 10 with (otherwise) random AR-parameters and uniformly distributed input. The noise is constructed of three independent sources, generated with 3 univariate AR-models with random parameters and uniformly distributed input, which were instantaneously mixed into the two sensors with a random mixing matrix. The relative strength of noise and signal was set randomly. The data were generated with this Matlab code (external link). Note, that the phrase 'simulated EEG data' is meant loosely. The simulation addresses the conceptual problems of EEG data, but e.g. the actual spectra can be quite different from real EEG data.

The task is to estimate the direction of the interaction of the signal. A submitted result is a vector with 1000 numbers having the values 1, -1, or 0. Here, 1 means direction is from first to second sensor, -1 means direction is from second to first sensor, and 0 means "I don't know".

For all examples either 1 or -1 is correct. The most important point here is the way it is counted: you get +1 point for each correct answer; you get -10 points for each wrong answer; and you get 0 points for each 0 in the result vector. With this counting confidence about the result is added into the evaluation. It is strongly recommended that for each example the evidence for a specific finding is assessed.


Submission


Please send submissions by October 10, 2011 to

Guido Nolte
email: guido.nolte(at)first.fraunhofer.de


Causal Analysis of simulated EEG data


Description of task


Motivation: Noninvasive electrophysiological measurements like EEG/MEG measure to large extent unknown superpositions of very many sources. Any relation observed between channels is dominated by meaningless mixtures of mainly independent sources. The question is how to observe and properly interpret true interactions in the presence of such strong confounders.

Download data (external link) here.

To read the data into MATLAB, type
fid=fopen('simuldata.bin');
data=reshape(fread(fid,'float'),6000,2,1000);

The data consists of 1000 examples of bivariate data for 6000 time points. Each example is a superposition of a signal (of interest) and noise. The signal is constructed from a unidirectional bivariate AR-model of order 10 with (otherwise) random AR-parameters and uniformly distributed input. The noise is constructed of three independent sources, generated with 3 univariate AR-models with random parameters and uniformly distributed input, which were instantaneously mixed into the two sensors with a random mixing matrix. The relative strength of noise and signal was set randomly. The data were generated with this Matlab code (external link). Note, that the phrase 'simulated EEG data' is meant loosely. The simulation addresses the conceptual problems of EEG data, but e.g. the actual spectra can be quite different from real EEG data.

The task is to estimate the direction of the interaction of the signal. A submitted result is a vector with 1000 numbers having the values 1, -1, or 0. Here, 1 means direction is from first to second sensor, -1 means direction is from second to first sensor, and 0 means "I don't know".

For all examples either 1 or -1 is correct. The most important point here is the way it is counted: you get +1 point for each correct answer; you get -10 points for each wrong answer; and you get 0 points for each 0 in the result vector. With this counting confidence about the result is added into the evaluation. It is strongly recommended that for each example the evidence for a specific finding is assessed.


Submission


The deadline for submission of results was October 31, 2011 to be sent to

Guido Nolte
email: guido.nolte(at)first.fraunhofer.de


In addition, each participant was asked to provide basic information about his/her algorithm (e.g. a bibliographical reference) .


Results


To see the algorithm details, click the submitter's name.

Name City Total points Correct detections False detections Details
A.M. Bianchi Milano/Italy -2289 701 299
S. Hu Hangzhou/China 252 352 10 Click name
L. Leistritz Jena/Germany -357 773 113 Click name
V. Vakorin Toronto/Canada 218 278 6 Click name
M. Wibral Frankfurt/Germany -247 163 41 (pdf) (external link)


Remark: The total points can be calculated as the number of correct detections minus ten times
the number of false detections.




History

Legend: v=view, c=compare, d=diff
Date UserEdit Comment Version Action
Fri 04 of Nov., 2011 05:18 CET admin   35
Current
 v
Fri 04 of Nov., 2011 05:04 CET admin   34  v  c  d  
Fri 04 of Nov., 2011 05:03 CET admin   33  v  c  d  
Fri 04 of Nov., 2011 04:59 CET admin   32  v  c  d  
Fri 04 of Nov., 2011 04:58 CET admin   31  v  c  d  
Fri 04 of Nov., 2011 04:55 CET admin   30  v  c  d  
Fri 04 of Nov., 2011 04:54 CET admin   29  v  c  d  
Wed 02 of Nov., 2011 11:13 CET admin   28  v  c  d  
Thu 20 of Oct., 2011 14:13 CEST admin   27  v  c  d  
Sat 16 of July, 2011 08:07 CEST admin   26  v  c  d  
Sat 16 of July, 2011 08:05 CEST admin   25  v  c  d  
Fri 15 of July, 2011 14:32 CEST admin   24  v  c  d  

Menu

Google Search

 
sisec2013.wiki.irisa.fr
WWW