2016-11-01

the bento-box MCMC sampler

In a low-research day I did get in a conversation with NYU student Neelang Parghi about an old idea (code-named “bento box”) I have had (with my MCMC collaborators) about an MCMC method that would identify multi-modality, split the space, and recursively sample simpler and simpler problems until it is sampling much smaller, simpler problems that are technically trivial. Parghi is going to try to make this happen as part of his Computational Physics class term project.

No comments:

Post a Comment