Interpreting Two Psychological Functions by A Hierarchically Structured Neural Memory Model
Tetsuya Roya · 2002
In this paper, a novel neural memory model is proposed. The neural memory model proposed in this paper is based upon generalized regression neural networks (GRNNs) which are the paradigms ofradial basis function neural networks (RBF-NNs). With the benefit of their quick learning capability and robustness, the application of GRNNs has been rapidly increased in many disciplines. Then, with in the context of a newly proposed hierarchically arranged generalized regression neural network (HA-GRNN), two psychological functions, intuition and attention, are interpreted in terms of the evolution of the HA-GRNN. Within the framework of HA-GRNN, two types of memory, namely both the long and short term memo ry motivated from biological and cognitive studies, are considered and a dynamic learning system is thus proposed. In the simulation study, the effectiveness of the HA-GRNN in comparison with k-means clustering method is confirmed within the context of pattern classification tasks.