Designer networks for time series processing

The conventional tapped-delay neural net may be analyzed using statistical methods and the results of such analysis can be applied to model optimization. The authors review and extend efforts to demonstrate the power of this strategy within time series processing. They attempt to design compact networks using the so-called optima brain damage (OBD) method. The benefits from compact architectures are three-fold. Their generalization ability is at least comparable,they involve less computational burden, and they are faster to adapt if the environment changes. It is shown that the generalization error of the network may be estimated, without extensive cross-validation, using a modification of Akaike's final prediction error (FPE) estimate (1969).<
>