MLP Train Methods
Page 1 of 1

Author:  jperezmarquez [ Tue Jan 26, 2021 10:10 am ]
Post subject:  MLP Train Methods

Concerning multilayerperceptron train Algorithms with commercial version.
Is mlptrainlbfgs better than mlptrainnetwork for high volume tasks (a million of samples and 200 hundreds variables)??
Does mlptrainlbfgs use multithreading (in c# version)? I set the option alglib.parallel at the end, but it seems that it only works with one core. Coefficients randomization works if I launch on my own parallel tasks for each restart?? Is the randomization seed shared between multiple tasks???
Is possible that mlptrainlbfgs starts training with the current coefficient values??? In mlptrainnetwork counterpat it is possible to put restarts=0 in order to take current coefficients...

Page 1 of 1 All times are UTC
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group