A new neural network model emphasizing importance for associative memory

The Hopfield network is one of the most important types of neural network, with its readiness for hardware implementation and successful applications in associative memory (AM). However, when used for AM, it has three drawbacks: low capacity, slow convergence speed and weakness. The higher order correlation network (HOCN), suggested by Y.C. Lee (1986), is a direct generalization of the principles which play an important role in the construction of the Hopfield network. It enhances the network´s ability by degrees if with a correlation order K (K>2). As for practical applications, unfortunately, difficulties arise due to the complexity in its hardware implementation. In this paper, based on some properties of the human memory, the authors have modified the Hopfield network and suggested a new neural network, emphasizing the importance for AM. It seems the new network has some advantages in its abilities and hardware implementation, so that it is better than both Hopfield´s and Y.C. Lee´s networks