Encoding Prior Knowledge into Data Driven Design of Interval Type-2 Fuzzy Logic Systems

In system identification or modeling problems, interval type-2 fuzzy logic systems (IT2FLSs), which have obvious advantages for handling different sources of uncertainties, are usually constructed only using the information from sample data. This paper tries to utilize the information from both sample data and prior knowledge to design IT2FLSs to compensate the insufficiency of the information from single knowledge source. First, sufficient conditions on the antecedent and consequent parameters of IT2FLSs are given to ensure that the prior knowledge can be incorporated into IT2FLSs and three kinds of prior knowledge – bounded range, symmetry (odd and even) and monotonicity (increasing and decreasing) – are explored. Then, design of IT2FLSs using the information from both sample data and prior knowledge is transformed to the constrained least squares optimization problem. At last, to show the superiority of the proposed method, simulations and comparisons are made.