Mutually linked HASPs: A solution for constraint-satisfaction problems by associative processing

A model of a human associative processor (HASP) is applied to a class of constraint-satisfaction problems which can be seen as a formal representation of cognitive tasks such as memory retrieval and interpretation of scenes and sentences. A problem consists of a set of nodes, a set of labels, and relational constraints. A relational constraint specifies a set of compatible node-label pairs, and a solution consists of a set of labels of all nodes which satisfy the constraints. The structure of the problem is represented by the associational structures defined on mutually linked HASPs, where a node corresponds to a single HASP and a solution consists of a set of output patterns from respective HASPs, all of which satisfy the relationships defined between them. Relational structure defined between HASPs are divided into three classes according to the difficulty in solving problems. A series of simulation studies has shown that mutually linked HASPs can solve problems very efficiently by associative and parallel processing capabilities.