slow independent component analysis, SICA algorithm
Code for a paper A new independent component analysis for time series feature extraction with the concurrent consideration of high-order statistic and slowness
SICA is a new algorithm for feature extraction wich conbines the advantages of SFA and FastICA.
SICA could sort the extracted features acoording to the slowest and extracted more useful features than FastICA.
And here are some reults in the paper
where the SICA could figure out the right orignal features but the FastICA can't
the SICA has been packed for convient use.
the are two differnent algorithm about SICA in the package:
"sica" for unsurpvised learning cases
"ssica" for surpvised learning cases
pip install SICA
from SICA import sica res = sica.SICA.fit_trainsform(traindata) from SICA import ssica traindata = ssica.fit_trainsform(traindata,trainlabel) testdata = ssica.trainsform(testdata)