Neural Network Validation Criterion Reached?

Hi, I am currently trying to adapt the example for a CNN to classify images of numbers to classify a dataset of fingerprint I have aquired. When I am training the neural network though it say "Validation Criterion Met" and ends the training even though accuracy is only around 20 percent or so and training for only 1 epoch. Why is this?

Also when trying to do the future prediction labels I get this error even though my GPU has 6GB of RAM. What do I do?

Error using nnet.internal.cnngpu.batchNormalizationForwardPredict Out of memory on device. To view more detail about available memory on the GPU, use 'gpuDevice()'. If the problem persists, reset the GPU by calling 'gpuDevice(1)'.

Error in nnet.internal.cnn.layer.util.BatchNormalizationGPUStrategy/forwardPredict (line 20) Z = nnet.internal.cnngpu.batchNormalizationForwardPredict(X, ...

Error in nnet.internal.cnn.layer.BatchNormalization/predict (line 126) Z = this.ExecutionStrategy.forwardPredict( ...

Error in nnet.internal.cnn.SeriesNetwork/activations (line 55) output = this.Layers{currentLayer}.predict( output );

Error in nnet.internal.cnn.SeriesNetwork/predict (line 37) output = activations(this, data, indexOutputLayer);

Error in SeriesNetwork>iPredictSingleNumericObserv (line 919) Y = gather(predictNetwork.predict(X));

Error in SeriesNetwork/predict (line 338) Y(:,:,:,i) = iPredictSingleNumericObserv(predictNetwork,X,GPUShouldBeUsed); Error in SeriesNetwork/classify (line 559) scores = this.predict( X, varargin{:} );

Answer 1

[Greg Heath]

QUESTION: When I am training the neural network though it say "Validation Criterion Met" and ends the training even though accuracy is only around 20 percent or so and training for only 1 epoch. Why is this?

ANSWER: The default Validation Criterion is 6 . Therefore if the error on the validation subset increases continuously for 6 epochs, training terminates.

Hope this helps.

Thank you for formally accepting my answer

Greg