Symbolic Knowledge Encoding Using a Dynamic Binding Mechanism and an Embedded Inference Mechanism
This chapter describes how synchronous activity between neuron elements can be used to build a dynamic binding mechanism and knowledge encoding mechanisms in a connectionist manner. The purpose of these mechanisms is to build a connectionist inference architecture that can replicate common symbolic styles of inference. To build such an inference architecture, an extended temporal synchrony approach (Park, Robertson, & Stenning, 1995) is used as a basic building block. This is a revision and extension of an approach to the dynamic binding problem in connectionist systems, proposed by Shastri and Ajjanagadde (1993). In addition, we introduce a set of algorithms that gives us a means of compiling a class of symbolic rules into a uniform inference network called a structured predicate network. This is used as a connectionist knowledge encoding mechanism which encodes symbolic rules and supports very fast inference.