Enhancing Connectionist Expert Systems by IAC Models through Real Cases

This work presents a study of learning (case-based) in an interactive activation and competition (IAC) connectionist model. In this type of neural network, the basic learning mode may be classified as rote learning, and no iterative algorithm is used. The knowledge elicitation corresponds directly to the connection weights and its values are obtained by a type of engineering called connectionengineering. In a way it is similar to the knowledge engineering in that it obtains functioning rules for an expert system. In this sense, an example of differential diagnosis in rheumatology is used to study the learning performance of a neural network with the introduction of real clinical cases, presented by a expert doctor. These clinical cases are used as a source of additional knowledge that represent relations between diseases and symptoms.