Simplistic MCMC ensemble sampler based on emcee, the MCMC hammer
Find a file
Adam Reichold 1a282ed1c7
All checks were successful
/ check (push) Successful in 21s
Fix CI badge to target Codeberg instead of GitHub.
2026-02-21 10:54:07 +01:00
.forgejo/workflows Migrate CI from GitHub to Forgejo. 2026-02-17 11:03:38 +01:00
src Remove spurious dependency of moves on models instead of just parameters 2025-12-23 12:34:54 +01:00
tests Make moves generic and add alternative and mixed moves. 2025-12-23 11:25:45 +01:00
.gitignore Drop the Python bindings as the efficiency gain does not offset the cost of calling into the single-threaded Python interpreter. 2021-12-21 14:23:57 +01:00
Cargo.toml Migrate CI from GitHub to Forgejo. 2026-02-17 11:03:38 +01:00
LICENSE-APACHE Add license and extend crate meta-data. 2021-12-21 12:42:32 +01:00
LICENSE-MIT Add license and extend crate meta-data. 2021-12-21 12:42:32 +01:00
README.md Fix CI badge to target Codeberg instead of GitHub. 2026-02-21 10:54:07 +01:00

hammer-and-sample

crates.io docs.rs codeberg.org

A simplistic MCMC sampler implementing the affine-invariant ensemble sampling of emcee with serial execution and optionally with parallel execution based on Rayon.

The implementation is relatively efficient, for example computing 1000 iterations of 100 walkers using the hierarchical model from hierarchical.rs takes approximately 1 min using emcee and multiprocessing versus 50 ms using this crate and Rayon, running on 8 hardware threads in both cases.