Research Of Teaching-learning-based Optimization Algorithm

Compared with traditional optimization algorithms,intelligent optimization algorithms have the advantages of simple principle,strong operability and ability to solve complex optimization problems.Therefore,intelligent optimization algorithms play an irreplaceable role in solving optimization problems.Numerous scholars have devoted themselves to the research of intelligent optimization algorithms.Indian scholars Rao etc.proposed an intelligent optimization algorithm in 2011 which mimics the process that teachers impart knowledge to students and students learn knowledge by themselves.Therefore,the algorithm is named teaching-learning-based optimization algorithm(TLBO),its principle is simple,there are few control parameters,and it has faster convergence speed.Since it was proposed,scholars have conducted widely and deeply research on it,and many improved algorithms have been proposed.Most of these algorithms have been widely applied to various fields and industries.The principle of TLBO is simple,but the local search capability is lacking,and it is easy to fall into a local optimal solution.This paper addresses these deficiencies and has done the following research work on solving unconstrained optimization and constrained optimization problems:(1)For the unconstrained optimization problems,a hybrid teaching-learning-based optimization algorithm is proposed.In order to enhance the search ability of the algorithm,the opposition-based learning strategy is integrated into the basic teaching-learning-based optimization algorithm.In every iteration the current point and the quasi-opposition point are evaluated respectively,the best individual is selected to join the next iteration.At the same time,it proposes a search strategy called quasi-gradient descent which guides the search direction and makes the evolution of the population more directional.At the end of the learning phase,a ring neighbor structure is added to driven individuals learn from students randomly select from the neighbors or all populations,which increased the local search ability and ensured population diversity.Through simulation analysis,the hybrid algorithm shows good performance in solving low-dimensional optimization problems and high-dimensional optimization problems.The accuracy and speed of the solutions are superior to most of the algorithms in this paper.(2)For the constrained optimization problems,a co-evolutionary teaching-learning-based optimization algorithm is proposed.The optimal solution of the optimization problem is obtained through the cooperation between two populations.The tolerance amounts from transforming equality constraints into inequality constraints are firstly added to the coevolutionary parameter set to realize adaptive evolution.A self-learning strategy of random mutation is proposed to increase the population search efficiency,and the fitness function of existing co-evolutionary algorithms is improved based on the statistics knowledge.The evolutionary process of population is divided into early stage and later stage.In different stages,the fitness functions are adjusted according to the different goals of two stages which makes the search process more objective and increases the population diversity.This algorithm ensures the excellent characteristics of teaching-learning-based optimization algorithm.Efficiency of the proposed algorithm is verified by solving equality constrained optimization problems and engineering problems.