Just wondering if there is any official documentation for the DNN methods in TMVA? Is it using CNN as background algorithm or something else? What’s the meaning of those parameters and options inside? Can it also be applied on distributed computing system?
Official documentation (updated TMVA users manual) is upcoming soon. The DNN can run on various distributed systems: multi-threaded cpu, GPU cuda, GPU OpenCL.
It’s another year but seems DNN is still not available in the current officual documentation? Anywhere I can get a pre-release version?
Hi Qin Gang,
Pre-release documentation with usage instructions for the DNN is available here: https://github.com/root-project/root/blob/master/documentation/tmva/UsersGuide/TMVAUsersGuide.pdf
Please note that the DNN can run using a multi-threaded CPU implementation and using the GPU currently.
The latest version of the TMVA guide has a chapter on DNN. Hopefully that helps.