Combined mutually connected neural network model for higher order association

A combined mutually connected neural network model for high-order association is proposed. It contains a plural functional modules, each of which is mutually connected to a neural network with hidden units in order to improve the recall performance. The model comprises three different type of blocks, i.e., sub-net, intensive-net and intensive sub-net. Each module works dynamically in cooperation with other functional modules. The higher-order association between three pieces of two-dimensional character dot patterns and corresponding three-character word patterns are demonstrated by the model