Adopt Frank Noé's suggestions

choderalab / msm-pipeline

A pipeline for MSMs.

GNU Lesser General Public License v3.0

2 stars 5 forks source link

Open maxentile opened 7 years ago

maxentile commented 7 years ago

@franknoe reviewed our pipeline and suggested a few changes:

[ ] Use commute-map weighting rather than kinetic-map weighting during tICA projection.
[ ] Instead of truncating always to 50 tICs, we can retain the components with timescales > the lagtime.
[ ] Try k-means instead of minibatch k-means. This may be more memory-intensive, but should be feasible if we stride our data before clustering.

sonyahanson commented 7 years ago

Sounds great!

sonyahanson commented 7 years ago

(updated your check boxes, assume that's what you were going for)