Skip to content

Optimal swap rate for HMMs

Home / Open Questions / Optimal swap rate for HMMs

Background: Optimal swap acceptance rates for parallel tempering algorithms are well-established for specific classes of distributions, but the theoretical optimal rate for hidden Markov models remains undefined.

Question / Future Work: The parallel tempering (PT) algorithm relies on an inverse-temperature schedule to facilitate exploration of multimodal posterior distributions. While a swap acceptance rate of 0.234 is often cited as near-optimal for certain high-dimensional problems, determining the specific swap acceptance rate that maximizes mixing efficiency for hidden Markov models (HMMs) is an unresolved theoretical challenge.

Why It Matters: Identifying the optimal swap acceptance rate is critical for minimizing computational costs and ensuring efficient exploration of high-dimensional, multimodal posterior distributions in Bayesian HMM inference.

Metadata & Links