TMVA regression warning

TMVA
1

Dear expert,

I’m using TMVA multi-regression, I got the following warning:

— MLP : ANN too complicated: #events=991 #synapses=1164

I find very little instruction about how to use regression, so I set the parameters to the ones used in tmva regression example:
factory->BookMethod( TMVA::Types::kMLP, “MLP”, “!H:!V:VarTransform=Norm:NeuronType=tanh:NCycles=2000:HiddenLayers=N+20:TestRate=6:TrainingMethod=BFGS:Sampling=0.3:SamplingEpoch=0.8:ConvergenceImprove=1e-6:ConvergenceTests=15:!UseRegulator” );

Could you please kindly let me know what does the warning mean and how to solve this problem?

Thanks,
Binghuan

2

Hi,

I think the warning is caused by the fact you are using a too complex NN architectures (too many nodes) compared to the small number of events you are having for training.
Either increase the number of training events, or simplify significantly the NN.

Best Regards

Lorenzo