I'm experimenting with training neural networks.

Map[NetTrain[NetInitialize@#, trdata] &, {net1, net2}]

shows the Training Progress monitor on each calculation. But

ParallelMap[NetTrain[NetInitialize@#, trdata] &, {net1, net2}]

doesn't show those.

Is it possible to forward the Training Process monitors to the front end in a parallel computation?

  • $\begingroup$ On my machine (11.0.1), NetTrain already uses all CPU cores. I would expect ParallelMap to result in reduced performance compared to a serial run. Why do you want to parallelize manually? Do you have any evidence that it helps? $\endgroup$ – Szabolcs Feb 22 '17 at 8:31
  • $\begingroup$ @Szabolcs I ran a test on my machine: singel train ~220s, 4 nets parallel ~400s each. It might be influenced by Hyperthreading and Turbo boost of my i7 processor. $\endgroup$ – meneken17 Mar 7 '17 at 13:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.