search

lukas-weber / Carlo.jl

Monte Carlo framework that provides MPI parallelization, checkpointing and statistical postprocessing in an algorithm-agnostic way.
MIT License
27 stars 2 forks source link

Direct sampling support #9

Closed hz-xiaxz closed 3 weeks ago

hz-xiaxz commented 2 months ago

Hello! This is quite a nice package! I'm a student working on a project using direct sampling method to simulated Fermion system, so it is different from the MCMC framework. But I'm still interested in the MPI frame and the job arrangement system this package. Will it be possible for me to take advantage of this package by e.g. set thermalization=0 and use the MPI frame and autocorrelation caculation? I'll be appreciate if any guidance could be given. Or is there existing package that support parallel caculation of direct sampling problem?

hz-xiaxz commented 2 months ago

If you are interested, maybe we can work on a PR to add this feature

lukas-weber commented 2 months ago

Hi, apologies for the late reply!

In a still unpublished code, I am actually using Carlo as the framework to simulate Fermion systems using the Auxiliary-field quantum Monte Carlo method myself. That method is also not MCMC, but a one-sided projection/random walk.

Basically, Carlo gives you a way to process random numbers in parallel. If your method has no thermalization time, you can set thermalization to zero and it should work. If there are additional features you need, we can talk further :)

hz-xiaxz commented 3 weeks ago

Thanks, setting thermalization to zero works all well for my code!