Training data plots are useful.
However, there is no indication of how good the net will perform on nontraining data (THE TRUE MEASURE OF GOODNESS!)
The best SINGLE measure of goodness is the ratio of mse(error) to sum(var(target',1)) for the test subset.. The latter is mse(error) for the reference: a naive constant output model
output = mean(target').
Details are available by searching
Thank you for formally accepting my answer