Standard backend DNN in TMVA

Dear all TMVA experts,

Right now I want to try the DNN model in TMVA package, and I saw in the TMVA user guide that there is the

The standard backend is the fall-back implementation and will work on any platform where Root can be installed. The CPU backend uses multithreading to perform the training in parallel on multi-core CPU architectures and requires a multithreaded BLAS implementation and the Intel TBB library. The GPU backend can be used to perform the training on CUDA-capable GPU architectures.

Besides the CPU and GPU multi-core DNN models, does there exist the standard DNN in the TMVA package, without installing the CPU and GPU library?

Kind regards,


The standard backend is currently disabled. The User’s guide is in the process of being updated with this information.


Dear Kim,

Got it! Sure I will stick with the CPU and GPU ones with the DNN after setting up the environment.


1 Like