Skip to content

ROADMAP.md

Latest commit

 

History

History
30 lines (20 loc) · 1.2 KB

ROADMAP.md

File metadata and controls

30 lines (20 loc) · 1.2 KB

Roadmap

This roadmap captures planned feature development and priorities. It is a living document and will evolve as the project grows.

Next (1–2 months)

  • TBD

Done

  • Support other interacting-particle variants (e.g., SVGD-like updates).
  • Benchmark suite with reproducible performance comparisons.
  • Paper-ready experiment scripts and dataset loaders.
  • Add example: hierarchical Gaussian model.
  • Add example: Bayesian logistic regression.
  • Add convergence diagnostics (ESS, R-hat-like proxies) and trajectory summaries.
  • Implement adaptive step-size and stability checks for Euler–Maruyama.
  • Expand documentation with a "Getting Started" tutorial and API reference pages.
  • Add a higher-level Sampler wrapper API with clear configuration, seeding, and diagnostics.
  • Improve MMD^2 support by adding kernels (RBF, Matérn) and bandwidth selection utilities.
  • Add PrO drift implementations for additional scoring rules (e.g., energy score, CRPS) with reference examples.

Near-Term (3–6 months)

  • Provide GPU-friendly paths via JAX (optional backend) while keeping NumPy-first default.

How to Contribute

If you'd like to propose or prioritize items, open an issue or PR with your suggestion.