Semantic constraints for membership function optimization

The optimization of fuzzy systems using bio-inspired strategics, such as neural network learning rules or evolutionary optimization techniques, is becoming more and more popular. In general, fuzzy systems optimized in such a way cannot provide a linguistic interpretation, preventing us from using one of their most interesting and useful features. This work addresses this difficulty and points out a set of constraints that when used within an optimization scheme obviate the subjective task of interpreting membership functions. To achieve this a comprehensive set of semantic properties that membership functions should have is postulated and discussed. These properties are translated in terms of nonlinear constraints that are coded within a given optimization scheme, such as backpropogation. Implementation issues and one example illustrating the importance of the proposed constraints are included.