msultan

Steps to do this - [ ] Get the likelihood "energy" - [ ] Use the maximum of the likelihood "energy" as the map estimate The map estimate is the...

As in a completely separate thing from msmbuilder? Sent from my iPhone > On Sep 18, 2016, at 11:31 AM, Carlos Hernández [email protected] wrote: > > Wanna make this a...

Started in #876

After talking with @cxhernandez and @mpharrigan yesterday, it seems like there is a way to make multiseq featurizers completely generic without needed to code up a new variation for each...

No this needs a lot more fleshing out. Let's keep this open for now. Sent from my iPhone > On Jan 11, 2017, at 4:29 AM, Brooke Husic wrote: >...

Move to 3.8, this is likely to take a while Sent from my iPhone > On Jan 11, 2017, at 11:31 AM, Brooke Husic wrote: > > @msultan, do you...

I can reproduce this on my machine too. I think this has something to do with @jadeshi 's error calculation code. @jadeshi do you wanna take a look?

@jadeshi does the error calculation require a sink to be specified? ``` python mfpts1 = tpt.mfpts(msm,errors=False) ``` works just fine . If so, maybe we should add that explicitly to...

> RawPositionsFeaturizer(atom_indices=([[18, 35])), ref_traj=None) That is incorrect, the correct way is ```python RawPositionsFeaturizer(atom_indices=np.arange(18, 36), ref_traj=None) ``` This is assuming you want to include atom 35. >If I just want CA...

@smsaladi I think the sampling assumes that the trajectories are a dictionary. This is inconsistent with the API which uses lists of lists. A fast solution is to make the...