Open palominohernandez opened 1 year ago
Hi, that's correct! Currently deeptime only contains native VAMP. It's probably pretty easy to implement a symmetrized version to enforce reversibility (same as for TICA), stochasticity constraints as in deep MSM and deep rev MSM are a bit more involved. An intermediate would be training a VAMPNet lobe and use that as feature function before applying TICA - potentially stacked with other (derived from eg intuition) feature functions.
Dear all,
I was having a look at the code for VampNets that is in deeptime, and it seems that it is the one from mardt2018vampnets, right? Thus, it does not implement the reversibility and stochasticity constraints from _Deep_revmsm, right? I just wanted to know if I am right.
Thanks a lot!