SUPERVISED LEARNING EXTENSIONS TO THE CLAM NETWORK

The contextual layered associative memory (CLAM) has been developed as a self-generating structure which implements a probabilistic encoding scheme. The training algorithms are geared towards the unsupervised g eneration of a layerable associative mapping (Thacker & Mayhew, 1989). We show here that the resulting structure will support layers which c an be trained to produce outputs that approximate conditional probabil ities of classification. Unsupervised and supervised learning algorith ms operate independently permitting the unsupervised representational layer to be developed before supervision is available. The system thus supports learning which is inherently more flexible than conventional node labelling schemes. (C) 1997 Elsevier Science Ltd. All rights res erved.