Open maxentile opened 9 years ago
http://people.brandeis.edu/~berkes/data/papers/BlasBerkWisk_NeurComp2006.pdf
Linear SFA is formally equivalent to tICA with time-delay one.
Section 5: can jointly diagonalize time-delayed correlation matrices with more than one time delay, yielding a more robust solution.
SFA using deep neural nets: http://www.cv-foundation.org/openaccess/content_cvpr_2014/papers/Sun_DL-SFA_Deeply-Learned_Slow_2014_CVPR_paper.pdf
http://people.brandeis.edu/~berkes/data/papers/BlasBerkWisk_NeurComp2006.pdf
Linear SFA is formally equivalent to tICA with time-delay one.