search

imbs-hl / ranger

A Fast Implementation of Random Forests
http://imbs-hl.github.io/ranger/
769 stars 193 forks source link

parallel execution of importance_pvalues #718

Open chrissy005 opened 4 months ago

chrissy005 commented 4 months ago

I have a dataset with 156065 rows.

I have used ranger to generate a random forest model.

Is it possible to run importance_pvalues function parallely on this random forest model and dataset as I need to run it with 1000 permutations(num.permutations)?

mnwright commented 4 months ago

It's not implemented in the package because we didn't want to have a second parallelization in addition to the C++ parallelization. For large datasets, there probably is no benefit of running the permutations in parallel because the individual ranger fits run in parallel. If you want to implement it anyway, it's not hard. Here is the loop that needs to be parallelized: https://github.com/imbs-hl/ranger/blob/24a24bf7034d70a894a472e2f895c09488aa04f9/R/importance.R#L150 Just use your favorite parallelization package instead of sapply there (either in the package or just do it manually outside of the package).