Connection admission contorl in ATM networks using neurla netowrks: research directions and issues in commercial exploitation

摘要

A number of attempts have been made recnetly (roughly since 1990) to implement connection dmission ocntorl (CAC) in ATM networks by means of neural networks.These attempts use various mehtods and have met with varyign levels of success.they all try to solve some very serious networkign porblems related to the inadequacies of conventional algorihtms computing.lGiven that there has been little or no adoption of neural netowkr oslutions for CAC within the commercial world to date,there are serious questions as to whther or not research scientists are solving the problems which ocnmercial R&D are experiencing.This paper discusses these issues and their relevance to current ocmmercial development.