Hello, I have a large EEG dataset for which I need to do Independent Component Analysis (ICA)? Any leads on how to do it?
20 views (last 30 days)
Show older comments
I have got EEG dataset (98765 X 25 matrix) for which I need to do ICA.
0 Comments
Answers (2)
Raunak Gupta
on 17 Sep 2019
Hi,
For Reducing the dataset exact Independent Component Analysis (ICA) doesn’t exist but you may use rica for doing the dimensionality reduction. The Reconstruction ICA work closely as ICA but an added feature of optimizing non-linear objective function that also include penalty while reconstructing the output. This can improve the convergence speed as compared to ICA and the features you get are globally orthogonal. Reconstruction ICA is supported in R2017a and later versions.
For Other Dimensionality reduction methods you may look Dimensionality Reduction and Feature Extraction.
0 Comments
Mandavi Gahlot
on 22 Sep 2019
1 Comment
Raunak Gupta
on 23 Sep 2019
Hi,
From the EEG Dataset, I can understand that each data-point contains 25 features. Out of that I am assuming that last feature is label so effectively there will be 24 features so by doing rica mentioned above the number of feature for each data point can be bring down to less than 24. That way the information about the labels will not lost since each datapoint will have less than 24 features and one label. You may try to seperate labels first and then apply rica on 98765 x 24 matrix so the dimensionality can be reduced.
Hope this helps.
See Also
Categories
Find more on EEG/MEG/ECoG in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!