IMPLEMENTING NONLINEAR ACTIVATION FUNCTIONS IN NEURAL NETWORK EMULATORS

被引：8

作者：

SAMMUT, KM

JONES, SR

机构：

[1] Parallel and Novel Architectures Group, Department of Electrical and Electronic Engineering, University of Nottingham, Nottingham

来源：

ELECTRONICS LETTERS | 1991年 / 27卷 / 12期

关键词：

NEURAL NETWORKS; DIGITAL SYSTEMS;

D O I：

10.1049/el:19910645

中图分类号：

TM [电工技术]; TN [电子技术、通信技术];

学科分类号：

0808 ; 0809 ;

摘要：

An alternative approach to implementing nonlinear activation functions for digital neural networks is presented. Unlike other methods, this approach has the advantage that it is processed by a common arithmetic unit already required for the network computations and it does not require the use of a look-up table.

引用

页码：1037 / 1038

页数：2