"Jamaa Ambarak" <firstname.lastname@example.org> wrote in message <email@example.com>... > "Jamaa Ambarak" <firstname.lastname@example.org> wrote in message <email@example.com>... > > "Greg Heath" <firstname.lastname@example.org> wrote in message <email@example.com>... > > > "Greg Heath" <firstname.lastname@example.org> wrote in message <email@example.com>... > > > > > > ---SNIP > > > > > > > [ net, tr, Y, E ] = train(net,pn,T); % Yn = sim(net,pn); E = T-Yn > > > > > > Correction: If you do not normalize T, remove the subscript "n" from Y. > > > > > > Greg > > please Greg tell me > > first I have tried to train the neural without normalize but the results are very bad because the data should be normalize (speed velocity is 20-80 mile per hours and the other big also ). I did not normalize T , but how can I convert y back to the original scale using > > y1 =mapminmax('reverse',y,ts);
Just use whatever is in the documentation.
> also how can I validate the training and testing
See comp.ai.neural-nets for the NN concepts of validation and testing.
data = design + test design = train + validate
Training set estimates weights
Validation set is used repeatedly with the training set to determine all other learning parameters (especially the stopping epoch), and to choose the best out of multiple designs.
Test set is used on 'best' design to obtain an unbiased estimate of performance on nondesign data.
If the test set performance is unsatisfactory, the data must be divided again before designing new candidates.