The Classes of Left and Right sigmoidal signals for Feed forward Neural Networks

摘要

The authors recently introduced left sigmoidal signals and right sigmoidal signals to prove certainfunction approximation theorems for Feed Forward Neural Networks. In this study we identify a class LS (resp.RS.) of left sigmoidal (resp.Right sigmoidal) signals having the property that the envelope of the derivativesof members of a class LS (resp. RS) is a left sigmoidal (resp.Right sigmoidal) signal.