Conventional and Connectionist Approaches to Face Processing by Computer
Phillips W & Smith L (1989) Conventional and Connectionist Approaches to Face Processing by Computer. In: Young A & Ellis H (eds.) Handbook of Research in Face Processing. Amsterdam: Elsevier, pp. 513-518. http://www.sciencedirect.com/science/book/9780444871435#ancsection10
In this brief commentary on Bruce and Burton's review of face processing by computer we concentrate on further emphasising the differences between approaches based upon conventional von Neumann computing, and those based upon neurally inspired computing. Our aim is to outline some of the main differences between the two approaches, and to show why neurocomputation is well suited to face processing. Empirical test of these arguments will be provided by the success or failure of applications of neurocomputing to face processing over the next few years. First, we discuss the goals of face processing. We agree with Marr (1982) and with Bruce and Burton that such considerations should be given high priority. However, we also believe that investigation of these issues should be combined with that of the mechanisms or resources that are available for achieving the goals; just as goals and abilities in nature evolve together.
|Place of publication||Amsterdam|
Professor Bill Phillips
Emeritus Professor, Psychology
Professor Leslie Smith
Emeritus Professor, Computing Science