Bayesian Approach to Intelligent Control and Its Relation to Fuzzy Control

In many application areas including economics, experts describe their knowledge by using imprecise (“fuzzy”) words from natural language. To design an automatic control system, it is therefore necessary to translate this knowledge into precise computer-understandable terms. To perform such a translation, a special semi-heuristic fuzzy methodology was designed. This methodology has been successfully applied to many practical problem, but its semi-heuristic character is a big obstacle to its use: without a theoretical justification, we are never 100% sure that this methodology will be successful in other applications as well. It is therefore desirable to come up with either a theoretical justification of exactly this methodology, or with a theoretically justified modification of this methodology. In this paper, we apply the Bayesian techniques to the above translation problem, and we analyze when the resulting methodology is identical to fuzzy techniques – and when it is different. 1 Formulation of the Problem Need for expert systems and intelligent control. In many application areas ranging from medicine to economics, there are experts who consistently make very good decision. In the ideal world, we should all be helped by these experts: • every patient should be seen by the world’s best specialist in treating the corresponding disease, • everyone’s investment strategy should be advised by the world’s best financial advisors, etc. Realistically, however, this is not possible: there are very few best experts, and it is not possible to use them for every single decision making process.