groupICA-python
Please refer to the project website at https://sweichwald.de/groupICA/. We kindly ask you to cite the accompanying article (see below), in case this package should prove useful for some work you are publishing.
Quick install
pip install groupICA
The developer documentation is available at https://sweichwald.de/groupICA-python.
This repository holds the source of the groupICA package which implements the groupICA algorithm presented in groupICA: Independent component analysis for grouped data by N Pfister*, S Weichwald*, P Bühlmann, B Schölkopf.
Furthermore, as a courtesy to other python users, this package contains implementations of
- uwedge, an approximate matrix joint diagonalisation algorithm described here, and
-
uwedgeICA
, which essentially—for the right choice oftimelag
parameters—amounts to an implementation of several second-order-statistics-based ICA algorithms such as SOBI/NSS-JD/NSS-TD-JD (please refer to the groupICA article mentioned above for more details on this),
which may be helpful in their own right independent of the grouped ICA.