bug about trainNetwork for Trainning Convolutional Neural Network for Regression

as in demo of Trainning Convolutional Neural Network for Regression,

"[XTrain,~,YTrain] = digitTrain4DArrayData; [XValidation,~,YValidation] = digitTest4DArrayData; .... "

the final train code is "net = trainNetwork(XTrain,YTrain,layers,options);" we can see that XTrain and YTrain are read-ahead. because the data size in the demo is small. but if we are on large scale data, so how to deal with "XTrain" and "YTrain" .

does trainNetwork support "XTrain" and "YTrain" read online for regression?

[Greg Heath]

So why the use of the term

   BUG

?

Greg

[Jack Xiao]

sorry for the term. but how to solve the problem? does trainNetwork support "XTrain" and "YTrain" read online for regression?

Answer 1

[Wentao Du]

Not sure about "read online".

For large datasets, you can create a datastore (see this example using imageDatastore): https://www.mathworks.com/help/nnet/examples/create-simple-deep-learning-network-for-classification.html

More on datastore:

https://www.mathworks.com/help/matlab/import_export/what-is-a-datastore.html