Bilkent ACM SIGART (Special Interest Group on Artificial Intelligence)
by Atacan Canduroglu
Based on the known or presumed biology of the human brain, Artificial Neural Networks (ANN's) were first proposed as a model of computation in the mid 1940's. As an alternative to the now traditional symbolic paradigm, neural network systems were less brittle and could learn automatically. Unfortunately, following the discovery of major limitations in their learning ability, research in the field all but stopped in the late 1960's and it is only recently that new learning algorithms capable of overcoming these restrictions have been discovered. As a result, ANN's are now being employed in numerous engineering applications. This talk will look at the various forms of neural network that currently exist. It is a first step in a research project aimed at deciding whether and how connectionist-like ANN's might offer a serious alternative to the symbolic view of mind.
Tuesday, 1st Dec. 1998
at 17:40 in EA502 (Eng. Building)