I use a 3 layers convolutional neural network as a classifier for a problem. How do I determine the classifier has a high bias? Should I decrease the number of hidden units?
