rsun@chaos.cs.brandeis.edu (Ron Sun) (08/06/90)
RE: discrete neural models From going through the proceedings of the recently held INNC-Paris conference, one can easily notice the surge of interests in neural networks of discrete input/output/states (for example, P-ROM, i-PROM, DN/PDN, and Aleksander's PLN, etc.), beyond a originally small group in the London area. I learned that people are seeking linkage between this approach and the more conventional approach with continuous activation and weights for all links, and comparing their performance. It seems to me, although I could be wrong, that this approach is gaining momemtum now, and, due to a number of interesting learning algorithms proposed recently, is becoming a viable alternative to the 'classical' neural network models. especially in learning and performing in discrete input/output domains. Yet, there has been few discussions and publications in mainstream neural net journals, especially in US. What is your view on this type of neural network models? --------------------------------------------------------------------- Some relevant references: I. Aleksander, "neural computing architecture" MIT Press 1989 D. Gorse & J. Tayler, Hardware realisable Learning Algorithms, INNC, 1990 E. Filho et al, A Goal-seeking Neuron for Boolean Neural Networks, INNC 1990 R. Sun, The discrete Neuronal Model, INNC, 1990 R. Sun, A Discrete Neural Network Model for Conceptual Representation and Reasoning, CogSci, 1989 R. Sun, Rules and Connectionism, INNC, 1990 and a lot more! ---------------------------------------------------------------------- These models have been applied to a wide variety of example domains, for example, XOR problem (Aleksander 1989, Sun 1990) Variable bindings in rule based systems (Sun 1989, etc.) Modeling biological neural nets (Sun et al 1989, etc) control problems (Tayler & Gorse 1990. ) Vision problems (Aleksander et al 1989) From: Ron Sun Brandeis University Computer Science Waltham, MA 02254 rsun@cs.brandeis.edu ----------------------------------------