Use AbstractMCMC interface

fauconab / PolygenicRiskScores.jl

Polygenic Risk Scoring in pure Julia

Other

10 stars 2 forks source link

Use AbstractMCMC interface #9

Open sethaxen opened 3 years ago

sethaxen commented 3 years ago

Instead of implementing MCMC itself, this package could use the AbstractMCMC.jl interface to do sampling. This would enable various MCMC flavors in different packages to be used for the sampling, e.g.

AdvancHMC.jl: HMC/NUTS
AdvancedMH.jl: Metropolis-Hastings
AdvancedPS.jl: particle sampling

AdvancedMCMC.jl also has functionality for sampling multiple chains in parallel using multiple threads or cores.

jpsamaroo commented 3 years ago

Agreed! Unfortunately neither @fauconab nor I are very familiar with MCMC; we ported this package mostly as-is from Python because we needed the core algorithm to run faster for our research. We would appreciate any help with figuring out how to generalize the MCMC so that Julia-native approaches could be tried out instead!