[R-sig-hpc] Parallel linear model
Mark Seligman
mseligman at rapidbiologics.com
Thu Aug 23 18:30:30 CEST 2012
The WideLM package was developed to deal specifically with this type
of problem. It computes regression coefficients and performs a
limited amount of ANOVA using a GPU.
Models are currently limited to a small number of main effects and
interactions. The design matrix is also limited by the GPU's memory
size - typically several GB. Execution time scales nicely with the
number of models tested, i.e., your 'n' below.
We developed it specifically to look for evidence of epistasis in
genomic data, but would like to make it more generally accessible.
Constructive criticism is most welcome.
- Mark Seligman
> I wonder if someone has experience with efficient ways of implicit
> parallel execution of (repeated) linear models (as in the
> non-parallel example below)? Any suggestions on which way to go?
>
> Patrik Waldmann
>
> pval<-c(1:n)
> for (i in 1:n){
> mod <- lm(y ~ x[,i])
> pval[i] <- summary(mod)$coefficients[2,4]
> }
More information about the R-sig-hpc
mailing list