Please also provide the following information in addition to your question:
- Operating System:R-3.5.0
- brms Version:
Hello,
I am developing bayesian modelling system using brms package I am just wondering if is there any way to speed up the brm function (like parallelizing through the cores of the computer) is taking a bit long but is also true that data has lenght of 4788 points and possible variables to include as inputs in the model are above 100. About the model I can tell that model is just a linear regression with no hierchachies.
Thank you.
Cristofer Tadrissi
You may want to use argument cores as explained in ?brm.
Hello Paul,
I am doing this:
ncores = detectCores()
rstan_options(auto_write = TRUE)
options(mc.cores = parallel::detectCores())
t0<-Sys.time()
fit1<-brm(formula = as.formula(form),data = model.env$BBDD[model.env$BBDD$USE==1,], iter = 2000, warmup = 200,prior = prior_object,
cores=ncores)
tfinal<-Sys.time()-t0
it reduced time processing a lot, but still I would like to reduce even more time processing of the model is there any another option that I can use?
That’s hard to tell without seeing your actual model.
Hello Paul, thanks for replying.
which details do you need about model? Maybe, I can show you some stuff but with some limitations as the information that I am managing belongs to a company. Anyway, the thing I am building must work for any model.
Is there any other package or function you might know that I can use for fit a linear regression with multiple variables? brm is working well for me (in results) but time processing is too much for fit a model as also the model is being optimized by some parameters (so if optimization has is 200 iterations - 1 iteration takes 20 mins with brm paralellized - it would be 4000mins).
Kind regards,
Cristofer