I am training a feedforward backprop network. Each training generates a new value for the errors and the Rs. The question is, is there really an indicator one can use to stop training and at that point use that model?

More Joseph Xavier Francisco Ribeiro's questions See All
Similar questions and discussions