I haven't obtained time improvements (sometimes it even takes longer) using a parfor loop in place of either of my for loops in the below code.
What i'd like is multiple networks to be being trained at the same time, so that I can run this code on a cluster, with each cluster running a different set of neural networks.
If I change to parfor Neurons = 1:n I can tell the parfor loop is working because "Neurons" no longer changes sequentially/monotonically, as expected of parallel loops. However, only one "Neural Network Training (nntraintool)" GUI loads, which makes me think that only one worker is being deployed? On top of this I am gaining no time improvements. Is it possible to train multiple neural networks at the same time using different matlab workers? How?
Many many thanks.
tic
for init_conds = 1:m
for no_neurons = 1:n
net = patternnet(no_neurons);
net = configure(net,inputs,targets);
net.IW{1,1} = ...
net.b{1,1} = ...
net.LW{2,1} = ...
net.b{2,1} = ...
[net, tr] = train(net,inputs,targets);
outputs = net(inputs);
outputall = [outputall, outputs];
end
end
toc