Neural computing: a new route to software reliability

Neural networks, parallel distributed processing, or connectionism has been the focus of a rising new paradigm in artificial intelligence with sometimes a claimed association with neuroanatomy. If a trained network is viewed simply as an implementation of a function, it has to be admitted that it is a nonconventional implementation and is obtained by a nonconventional route. Networks may then be viewed as a radically different implementation of software engineering problems. In this paper the authors illustrate the potential for this viewpoint with the results of studies currently being pursued on the use of neural computing as an emergent technology for the software engineer. It is shown that `network programming´ constitutes a new programming paradigm for the software engineer, and it makes drastic changes to the traditional area on which it impacts-in this case software production. They demonstrate how certain software-engineering problems can be reliably implemented as multiversion, neural-network software systems