Neural optimization of linguistic variables and membership functions

Algorithms for extracting logical rules from data that contains real-valued components require the determination of linguistic variables or membership functions. Context-dependent membership functions for crisp and fuzzy linguistic variables are introduced, and methods for their determination are described. The methodology for the extraction, optimization and application of sets of logical rules is described. Gaussian measurement uncertainties are assumed during the application of crisp logical rules, leading to "soft trapezoidal" membership functions, enabling the optimization of linguistic variables using gradient procedures. Applications to benchmark and real-life problems yield very good results.