Recurrent Expert Networks

Proceedings of the AAAI Conference on Artificial Intelligence, 13

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 13

SIGART/AAAI Doctoral Consortium Abstracts

Research has shown that computational techniques such as neural networks often provide classification abilities that are more accurate than methods which rely on explicit knowledge acquisition alone. On the other hand, because no "reason" for a particular classification can be given when a computational technique has been used, human experts tend to be skeptical of such systems. As a result, many researchers have developed tools, called hybrid systems, which combine the pattern recognition capabilities and parallel processing of neural systems while retaining the domain knowledge encoded in expert systems (Medsker 1994).