Dynamic Neuro-symbolic Expert System
Neuro-symbolic, hybrid intelligent system, hybrid knowledge representation
Intelligent systems (IS) are types of expert systems that can interact and learn about the users who use it, or try to understand about your interests. The IS term encompasses both classical artificial intelligence, based on symbolic rules, as computational intelligence, incorporating fuzzy systems, artificial neural networks (ANN) and evolutionary systems. Symbolic intelligent systems have a set of initial rules which are continuously analyzed, generating decisions from them. Although simpler to understand, these systems have difficulties in dealing with inaccurate information, or that were not foreseen in the initial set of rules. On the other hand, models based on connectionist, for example, ANN, are extremely effective in completing patterns that are not clear, though, have a great disadvantage due to the high level of abstraction encapsulates the knowledge of how response was obtained. These two types of systems can be combined to overcome the disadvantages presented in each one, thus producing a powerful tool in industry. This article aims to describe an architecture of a hybrid computer system, combining the main paradigms of artificial intelligence (AI): symbolism and connectionism. The model will be able to encode the basic rules of an expert (based on propositional logic) in a neural network and from it infer new rules according to the use of the system, thus serving aid for decision contained in a decision of the operator.