kazewong / flowMC

Normalizing-flow enhanced sampling package for probabilistic inference in Jax
https://flowmc.readthedocs.io/en/main/
MIT License
200 stars 23 forks source link

Implement mix discrete and continuous sampling #187

Open kazewong opened 1 month ago

kazewong commented 1 month ago

In some use cases, the users may want to switch between discrete sampling and continuous sampling.

There are two parts to this problem:

  1. Let the user defining an objective function that support both continuous and discrete variables.
  2. Sample the variables.

The second step can be done with some thing like https://num.pyro.ai/en/latest/_modules/numpyro/infer/hmc_gibbs.html#DiscreteHMCGibbs