This repository contains an implementation of likelihood-tempered SMC based on Algorithm 17.1 of [1] (link here). Adaptive temperature selection is performed by the procedure outlined in
Please read the example notebooks to get started, first Gaussian.ipynb followed by TwoMoons.ipynb. The implementation requires the user to specify three main functions: log_prior, log_target, and proposal. Our implementations use either Metropolis-Hastings random-walk (MHRW) or Metropolis-adjusted Langevin (MAL) dynamics, always with Gaussian noise.
[1] Chopin, Nicolas and Papaspiliopoulos, Omiros. An Introduction to Sequential Monte Carlo. Springer, 2020.