Intermodular connection suitable for module addition

We propose a new intermodular connections model of multimodular associative neural networks ν-connection, which is suitable for module addition. A module of multimodular associative neural networks called MuNet, which can memorize and associate the patterns combined complexly, was presented by Ohsumi et al. (1993). However, in the case that some modules are added, MuNet needs to relearn all connection weights between modules, and has high computational complexity to relearn. Thus we propose the ν-connection which needs to learn the only connection weights between existing and additional modules. Our model needs low computational complexity and its additional learning is faster than relearning all patterns. We use the merits of RBF (Radial Basis Function) nets, and improve learning speed and other properties