FTDNNFocused Time-Delay Neural Network
References in periodicals archive ?
This network is well-suited to time-series prediction and will be thus adopted and named FTDNN.
For both the FTDNN and the NARX models, the tapped line length was bracketed around the first zero of the autocovariance function of the dry-bulb temperature of [T.sub.db] 11 h, as reported by Dodier and Henze (2004).
It is interesting to note that the FTDNN (13) is consistently inferior.