This paper proposes a neural network model, which generalizes McCulloch-Pitts' model (McCulloch and Pitts, 1943) and Hopfield's model (Hopfield, 1982, 1984). We prove that the generalized model converges and has all the desirable properties of both McCulloch-Pitts' model and Hopfield's model. It is easier and more natural to formulate some application problems with the proposed class of models than with McCulloch-Pitts' or Hopfield's models, e.g., the 3-Satisfiability problem. The continuous counterpart of the binary model is provided too. Also, single-attributed neuron model is extended to multi-attributed neuron model in both binary and continuous cases.
ASJC Scopus subject areas
- Artificial Intelligence