Hi, I am using a TMultiLayerPerceptron for data analysis, and I noticed training slows significantly with large tree sizes. Is this behavior expected? Any suggestions for perhaps incorporating other neural net software with ROOT?
Kevin
Hi Kevin,
large tree sizes as in many events you loop over, or as in many branches, of which you only use a subset? The latter can be avoided by calling data->SetBranchStatus() on the data and testing tree (see root.cern.ch/root/html/TTree#TTr … anchStatus). If your tree has many events then please let me know and I’ll try to reproduce it.
Cheers, Axel.