nnstart (neural network toolbox in Matab) 

To support : www.paypal.com...
Function fitting (nftool)
Pattern recognition (nprtool)
Data clustering (nctool)
Time-series analysis (ntstool)
Training algorithm
Levenberg-Marquardt (trainlm) is recommended for most problems.
Baysian Regularization (trainbr) can take longer but obtains a better solution for some noisy and small problems.
Scaled Conjugate Gradient (trainscg) is recommended for larg problems, as it uses gradient calculations, which are memory efficient than the Jacobian calculations the other two algorithms use.
Mean Squared Error (MSE) is the average squared difference between the outputs and targets. Lower values are indicative of better results. Zero means no error.
The regression R value measure the correlation between the outputs and targets. An R value of 1 means a close relationship, and 0 means a random relationship.
Error Histogram: a normal distribution indicate a good results.
Regression plot : for perfect fit, the data should fall along 45-degree line, where the network outputs are equal to the targets.