Control of memory search by matching features with presynaptic inhibition

The neural networks of an associative memory search the patterns with direct relation between input and output patterns. Therefore, they can not search when there is no direct relation between desired conditions and output patterns. For such cases, search with chaos was proposed. This search accesses the memory with dynamic linkage in the low dimensional space spanned by the simple features on target memory and with one-to-many correspondence. It is reported that chaotic neural networks generate the chaos easily, with appropriate parameters. The purpose of this paper is to realize the search with chaos in neural network, by applying this search to chaotic neural networks and by using the model of presynaptic inhibition to control the chaos. Further, three types of functions are examined for presynaptic inhibition.