Using Parallel Algorithm to Speedup the Rules Learning Process of a Type-2 Fuzzy Logic System

Since a type-2 fuzzy logic system (T2FLS) needs to perform type-reduction calculation, and has more parameters compare to a type-1 fuzzy logic system, the rules learning process of a T2FLS using a serial algorithm is relatively time-consuming. In this paper, a data parallel method is designed to construct a parallel algorithm for the rules learning of a T2FLS. Numerical experiments show that the proposed parallel algorithm is faster than a serial one, and also gets better performance. Furthermore, to simplify the process of deriving the formula of parameter-updating for a FLS by using error back-propagation (BP), a program package for rules learning of a FLS based on Tensorflow’s automatic differentiation function in Python environment is provided. One can obtain it for free from the github address: https://github.com/wangjiaw123/wjwRepository_test.