MEADS

tags: Markov Chain Monte Carlo

The kernel is a combination of:

Generalized HMC
Radford Neal's trick for persistent MH updates

That they call Persistent-MH Generalized HMC

ECA adaptation scheme for the parameters

Persistent-MH Generalized HMC

Ensemble chain adaptation

Heuristics for parameter adaptation

Pre-conditioning

Step size

Damping factor

Slice drift

Links to this note