search

markovmodel / deeptime

Deep learning meets molecular dynamics.
GNU Lesser General Public License v3.0
173 stars 39 forks source link

A question about whitening/centering transformations #37

Closed gongshuai0606 closed 1 year ago

gongshuai0606 commented 1 year ago

Hi, it seems that the line 340 x = x.mm(self.mul) of deeptime/time-lagged-autoencoder/tae/utils.py should bex = self.mul.mm(x) according to Zero-phase Component Analysis Whitening formula Z=VD^(−1/2)V^TX

clonker commented 1 year ago

possibly, however this project is not actively maintained anymore. i currently don't have the time to look into it, but you may check out https://github.com/deeptime-ml/deeptime .

YunruiQIU commented 1 year ago

Not really, that depends on the shape of input trajectories/ featurized data, obviously, the codes of time-lagged autencoder codes are based on data which has shape (length, dimensions of features/coordinates). So original codes should be right.

gongshuai0606 commented 1 year ago

Not really, that depends on the shape of input trajectories/ featurized data, obviously, the codes of time-lagged autencoder codes are based on data which has shape (length, dimensions of features/coordinates). So original codes should be right.

Thank you very much. I understand now