Netlab experts - how to validate a neural network?

Hi there,

I've used the netlab toolbox to train a neural network base on 5000 training points I also have 5000 validation and 5000 test points. I've trained the network, up to a point where the error the netopt algorithm displays at each iteration is only decreasing by tiny amounts (I believe this is the MSE?), and visually the training data and predicted output are very similar on a plot.

How do I go about quantifiably validating my data? I realise I need to run my validation set through the network, but how do I use the results to validate it? RMSE? And if so at what point is it low enough?

Netlab doesn't seem to have any validation functions, so I'm a bit frustrated.

Many thanks for any help!