Novel Teaching-learning-based Optimization Algorithm for Design of Digital Filters

Abstract

An improved teaching-learning-based optimization algorithm named NTLBO is proposed for IIR digital design in this paper. Conventional mathematical methods have failed when reduced order adaptive models were used for the purposes of identification of problems. NTLBO utilizes a multi-learning strategy and opposition learning to overcome this disadvantage of the basic TLBO. The quasi-opposition-based learning has been applied to increase the diversity of solutions and broaden the search space to improve the global search ability. The multi-learning strategy makes local search more effective so as to speed up the convergence. To make a tradeoff between exploration and exploitation properly, the teaching factor is redesigned to increase the likelihood of solutions jumping out of local optima. Experiments are carried out on the classical examples and comparisons are made as well. The results indicate that the NTLBO algorithm achieved preferable performance in both reduced and same order models of IIR digital filters.