Feature request: Stochastic Gradient-MCMC support (SGLD, SGHMC, etc.)
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 2.8k
- Forks
- 315
- Avg merge
- 3d 9h
- Merged PRs (30d)
- 27
Description
Issue #950 asked about SG-MCMC, and while BlackJAX has kernels (link), using them in NumPyro requires extra glue code for making the log density, transforms, minibatching, and diagnostics (I've been struggling with this a bit, hence my post). Native support would make it much easier to run SG-MCMC directly on NumPyro model---and I think it is a common enough setting that would be worth it. This is valuable not just for minibatching large datasets but also for models with randomized approximate likelihoods (e.g. EnKF, PF; 1901.10568, 1907.06986). Would others be interested in collaborating on adding this?
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reviewing the BlackJAX sgmcmc kernels linked in the issue and how NumPyro currently handles model log density, transforms, minibatching, and diagnostics. Done should mean native NumPyro support for running SG-MCMC methods such as SGLD or SGHMC without the extra glue code described here.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- numpy, python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Active
- Clarity
- Needs clarification
- Newbie friendliness
- 35/100