MATLAB: Neural network: warm-start simulation

Deep Learning Toolboxneural networks

Hi, profiling the simulation of a generic neural network I noticed that in the initialization part a great amount of time is spent in the net=obj2struct(net) function. Being sim the most time-consuming function in the program I'm writing, I was wondering if there's a way (beside writing my own sim function) to save this struct after the first time in order to speedup the next sim calls skipping the conversion in a "convert once, run multiple times" fashion.

Best Answer

See my answer in the Newsgroup.

Hope this helps.

Greg

Related Solutions

MATLAB: How to train a feedfordward neural network with error weights

 Use the command window
 help mse
 doc mse

Hope this helps.

Thank you for formally accepting my answer

Greg

MATLAB: Convolutional neural network: parallel layers

In Matlab 2017a this kind of CNN is not possible to configure, however, with the new release of the Matlab 2017b, new functions were developed for that cases, called connectLayers yers and addLayers.

Hope it helps,

Regards, Javier

Best Answer

Related Solutions

MATLAB: How to train a feedfordward neural network with error weights

MATLAB: Convolutional neural network: parallel layers

Related Question