Think!: A unified numerical–symbolic knowledge representation scheme and reasoning system

AbstractMore and more applications of artificial intelligence technologies are made in biomedical software and equipment. These applications are multiple: intelligent alarms, intelligent monitoring, diagnosis support,…. Several different knowledge representation schemes are already in use: decision trees, first-order logic expert systems, calculations (mathematical modeling), trained neural network simulations,…. All these techniques have their own preferred field of application, and they do not overlap. Building a complete diagnosis support tool would require the use of several of these techniques. The problem is therefore communication between these very different systems and the complexity of the composite result. This paper describes the Think! formalism: a unified symbolic-connectionist representation scheme which tries to subsume some of the precited formalisms. Being able to integrate these knowledge representation schemes in a single model enables us to use existing knowledge bases and existing knowledge extraction techniques to make them communicate and work together