Chainsai:- facilitating sampling of multimodal probability distributions
summary
Chainsail is an open source Python web service that implements Replica Exchange (parallel tempering) to sample truly multimodal posteriors, interoperates with PyMC and Stan, and automatically finds temperature schedules using a density‑of‑states based tuner. It runs chains in parallel on Kubernetes or locally, uses OpenMPI, Flask, Postgres, React and Nix, and scales on demand in the cloud. The system is built and working on examples but has limited adoption, some bitrot, and currently lacks NUTS and advanced samplers.
AI-written summary · about
chapters (8)
- 0:00Chainsai project intro
- 1:25Multimodal probability distributions
- 7:29MCMC with multimodality
- 11:20How Chainsai solves it
- 15:17Bayesian clustering example
- 20:59Why sampling all modes
- 21:20Current status and future
- 23:06Q&A
related talks
39mSaaaS - Sampling as an Algorithm Service
29mMotivating Probabilistic Programming
47mThe State of the Art for Probabilistic Programming
01h 26mProbabilistic Programming and Bayesian Computing with PyMC
36mReal-Time ML Pipelines: Feature Chaining with Chronon
01h 29mHierarchical Probabilistic Modelling in Real Life
01h 27mProbabilistic Python: An Introduction to Bayesian Modeling with PyMC
42mSolving Real-World Business Problems with Bayesian Modeling