What we did for monitoring previously was adding the listener, which outputs the training score of the model after each iteration:
MultiLayerNetwork model = new MultiLayerNetwork(network.build())ScoreIterationListener scoreListener = new ScoreIterationListener(1);model.setListeners(scoreListener);
This will give you some idea of the performance of the model, but only on the training data, but we typically need more than that--at least the performance on the validation set would be useful to know to see if we start overfitting or not.
So, let's read the validation dataset:
DataSetIterator valSet = datasetIterator(valUris);valSet.setPreProcessor(preprocessor);
For training, previously we just took the dataset ...
