Hi people, the past two weeks I've been working with the NN toolbox from matlab. I have 4 inputs which I'm training to approximate a target, the thing is that once I trained the network and plot the regression, I have a good regression index for the training set, above .9, but it's very poor for the validation and test sets, .13 and .20 respectively.

It's a 4-20-1 network, my goal is set to 1e-3, mu set to 0.1. If someone has a suggestion please let me know. Thanks.

Happy holidays!!!