Improving teaching–learning-based-optimization algorithm by a distance-fitness learning strategy

Abstract

The teaching–learning-based optimization (TLBO) algorithm, composed of a teacher phase and a learner phase, is one of the most popular global optimization approaches. It is inevitable for TLBO to suffer from premature convergence and entrapment in local optima when dealing with complex optimization problems. To solve this problem, a novel TLBO variant named distance-fitness learning TLBO (DFL-TLBO), which employs a brand new distance-fitness learning (DFL) strategy to enhance searchability, is proposed. The DFL strategy guides the whole population to learn from the teacher and a distance-fitness optimum (DFO) candidate simultaneously according to a differential operator to improve the exploitation and exploration capability, respectively. Experimental results on the tests of CEC 2014 reveal that DFL successfully improves the exploitation and exploration capability of TLBO in balance. The statistical test results and the convergence analysis results indicate that the DFL-TLBO algorithm ranks best and is significantly better than other compared algorithms in solving global optimization problems.