Matlab Neural network Mini-batch training

I am now starting to work with neural network toolbox in matlab for big data And I have quite a big dataset of medical image patches.

Currently training is quite slow and I want to use stochastic(feed the net one example at a time)/mini-batch(feed the network few examples at a time) training.

So : Is there a way to use mini-batch learning in matlab?

I found the "adapt" function however in their documentation matlab doesn't recommend using it, The optimal would be some form of training function that would automatically split my data to batches,

I did try several calls to batch on my own with 1 epoch however it didnt get as good results as a normal autoencoder

autotest = configAutoEncoder( 100 , 'epochs' , 1 , 'L2WeightRegularization' , .004 ,'sparsityRegularization' , 4 , 'sparsity' , .15);
autotest.trainParam.showWindow = 0;
epochs = 100;
batchSize = 5000;
batchStart = 1:batchSize:size(xTrain,2)-batchSize+1;
idx = randperm(size(xTrain,2));
idx = 1:size(xTrain,2);
pr = nan(epochs,length(batchStart));
for epoch = 1:epochs
    for k = 1:length(batchStart)
        bs = batchStart(k);
        bidx = idx(bs:bs+batchSize-1);
        [ autotest , tr ] = train( autotest , ...
            xTrain(:,bidx) , xTrain(:,bidx) ) ;
        pr(epoch,k) = tr.perf(end);
    end
    fprintf( 'epoch:%d , performence : %.2f\n' , epoch , pr(epoch,end) ) ;
end

I am using matlab 2015a Thanks,

[Ekta Prashnani]

I have the same question. Can minibatch/stochastic training be done using neural network toolbox?

[peterukk]

I also want to know this! Please help! I have been digging and digging but could not find an answer..there must be some way to apply minibatch training using e.g. the SCG training function (in my case)