A stigmergic approach to teaching-learning-based optimization for continuous domains

Abstract

Teaching-Learning-Based Optimization (TLBO) has shown considerable success in solving complex optimization problems by imitating the higher cognitive leader-follower interactions among students and teachers in a classroom. Here, we propose to solve function optimization with a stigmergic probabilistic memory structure based on the indirect cooperation and distributed information exchange among students of a class along with a diversity preserving operator. The stigmergic approach integrates the historical experience, while the probabilistic representation and the diversity preserving operators avoid premature convergence. The resulting Stigmergic Real-domain TLBO (S R TLBO) is applied to optimization problems with continuous variables. The proposed S R TLBO is evaluated on 40 standard benchmark functions, as well as in modeling the earthquake source and wave propagation in the East of Iran from actual data. Results indicate overall improved solutions of the S R TLBO algorithm when compared to several competing algorithms. In several cases, particularly those functions with a simpler structure, considerable improvement is also observed in the rate of convergence to the optimum solution.