A practical filter for conflicting or subjective data

There are many real world applications of computational intelligence which need to process conflicting or subjective data. In particular, links between experts and routine users of intelligent virtual reality systems (IVRS) can be established by analyzing the reactions and opinions of the experts. The analysis should produce a mathematical realization of the expert´s knowledge or judgment which can easily be incorporated into IVRS. Mechanisms suggested for approaching this problem are threefold: neural network clustering, fuzzy risk analysis for evaluation, and invoking programs which “memorize” an expert´s analytic habits. For control applications, cloning an expert can initialize a process, tracking adherence to non stationary set points can monitor system state variable shifts, and fuzzy risk analysis can encourage tuning the system to approach sensible performance measures. Examples of analysis of olfactory neurons in the rat and health monitoring of a missile guidance system are presented to illustrate these suggestions. The use of subjectivity can be a strong modeling tool when it is properly labeled and invoked. To further the appropriate use of subjectivity in analysis, expert decisions and analytic techniques are being monitored by a “memorizing” program for future analysis and extensions into virtuality