For a given MCMC sequence we want to obtain statistics, snapshots (set of specified indices) and log properties (e.g. density) for each MCMC sampler. The multistate sampler needs its reporter to provide the potential energies and state indices to calculate free energies. Additionally, the multistate sampler needs to save regular checkpoint files from which it is possible to restart the MCMC chain.
Note: This PR also contains the improved PRNG implementation. Now, the random number stream is a broadcaster in the sampler state of each MCMC move. This was necessary to ensure that random number streams were separated, which became an issue during this PR since trajectories were synchronizing.
Description
For a given MCMC sequence we want to obtain statistics, snapshots (set of specified indices) and log properties (e.g. density) for each MCMC sampler. The multistate sampler needs its reporter to provide the potential energies and state indices to calculate free energies. Additionally, the multistate sampler needs to save regular checkpoint files from which it is possible to restart the MCMC chain.
Note: This PR also contains the improved PRNG implementation. Now, the random number stream is a broadcaster in the sampler state of each MCMC move. This was necessary to ensure that random number streams were separated, which became an issue during this PR since trajectories were synchronizing.
Todos
Status