Skip to content

Optimizing Mutation Proposals and Dynamics

Home / Open Questions / Optimizing Mutation Proposals and Dynamics

Background: The convergence efficiency of CTMCMC algorithms is highly dependent on the design of the mutation proposal distributions and the specific dynamics governing transitions between states.

Question / Future Work: Investigating adaptive proposal mechanisms—such as jumps in dimension greater than one, HMC-based deterministic drifts, and advanced Gibbs/Collapsed Gibbs samplers—is essential to reduce autocorrelation and scale to complex Bayesian problems.

Why It Matters: Improving proposal schemes is critical for the practical scalability of CTMCMC to high-dimensional parameter spaces with unknown dimensionality.

Metadata & Links