Is there anyway to parallel training random forest.

opencv / opencv_contrib

Repository for OpenCV's extra modules

Apache License 2.0

9.36k stars 5.75k forks source link

Is there anyway to parallel training random forest. #3653

Open huangynn opened 6 years ago

huangynn commented 6 years ago

System information (version)

OpenCV => 3.1
Operating System / Platform => mac
Compiler => clang

Detailed description

Steps to reproduce

Random forest training use single thread, which is not time efficient. Is there anyway to parallel training random forest.

StevenPuttemans commented 6 years ago

I guess that the for loop right here allows for applying a parallel for. We only need to keep track somehow of which features are already taken by the process in previous trees. This can be done with a global lookup table.

huangynn commented 6 years ago

Is there any plan to add this feature?

danielenricocahall commented 5 years ago

I couldn't parallelize the tree loop, but I was able to parallelize the inner variable loop using parallelfor. On the unit tests, it results in ~2x speedup for regression - maybe I'll try to clean things up and create a PR.