Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
1710575 | Applied Mathematics Letters | 2006 | 6 Pages |
Abstract
In this work we consider a general class of continuous activation functions which may be neither bounded nor differentiable; however, many sigmoidal functions are included as special cases. With this class of activation functions we give a result on asymptotic stability for neural networks under a weak condition of nonnegative definiteness. Then we show that differentiability is a condition for its exponential stability.
Related Topics
Physical Sciences and Engineering
Engineering
Computational Mechanics
Authors
Weinian Zhang,