How should we implement two-stage SAMS weight adaptation?

choderalab / perses

Experiments with expanded ensembles to explore chemical space

http://perses.readthedocs.io

MIT License

181 stars 51 forks source link

How should we implement two-stage SAMS weight adaptation? #240

Open jchodera opened 8 years ago

jchodera commented 8 years ago

The one-stage procedure reduces the gamma prefactor too quickly to be practical, so we should switch to a two-stage scheme.

In the recommended two-stage scheme, we wait until all empirical probabilities are within 20% relative error of their target probabilities. We don't currently have a way to retrieve all possible chemical state keys from the ProposalEngine, so how should we handle the decision of when to switch to the second stage?

jchodera commented 8 years ago

This is related to issues we're dealing with in http://github.com/choderalab/sams