Improvement And Simulation Research On Teaching-learning-based Optimization Algorithm

Under the background of the information age,there are many complex optimization problems in the region of scientific research and engineering practice that have been greatly developed.How to design more effective optimization methods has become a hot spot in many academic fields.Inspired by the laws of natural phenomena and the intelligence of biological groups,intelligent optimization algorithms have been proposed,and they performed well in solving complex optimization problems such as multi-peak,non-differentiable,inseparable,and discontinuous problems.These algorithms have received extensive attention from scholars in the world,and a lot of new algorithm was proposed in succession.In 2011,based on school teaching principles,Rao,an Indian researcher,proposed a new intelligent optimization algorithm,teaching-learning-based optimization algorithm(TLBO),which compares students' performance to fitness values,and abstract the process of teacher teaching and students learning into the teaching stage and learning stage.Its principle is simple,it is easy to understand,and there are few control parameters.So after the algorithm was put forward,it had attracted the attention and interest of many scholars.At present,some problems are still existed in this algorithm.For example,compared with other algorithms,the basic TLBO has low accuracy,slow convergence speed and some lack of local search ability.In order to make up for these weaknesses,the following improvements are made in this paper:(1)According to the characteristics of strong exploration ability but weak exploitation ability of teaching-learning-based optimization algorithm,a new algorithm based on self-study mechanism is proposed with the general opposition-learning methods.In this algorithm,in order to give full use of teacher,which is the leading role in the evolution,the teacher individual can improve itself according to general opposition-learning methods,so as to effectively avoid the population falling into local optimization.At the same time,to enhance the diversity of the population,three learning methods are set up at the learning stage,which can make the students to freely choose their favorite ways with probability.The simulation results on 18 standard test functions show that the algorithm is not easy to track into local optimum and has good optimization performance in precision and stability.(2)In order to further improve the exploitation ability of TLBO,combined with clustering partition method a multi-classes interaction one is proposed.This algorithm uses a new method of clustering partition based on Euclidean distance to divide the initial population into several subgroups so that the neighborhood information can be used more effectively.At the same time,to enhance the connection and ensure the simultaneous evolution of various subgroups,after the teaching stage,the worst individuals among each subgroup learn from the optimal individuals of the population,after the learning stage,random individuals learn from other subgroups,so to further strengthen the population diversity and improve the precision.Set up the numerical experiments on 6 unconstrained functions,4 constrained functions and an engineer optimization problem named tension spring optimization problems,and the simulation results show that compared with other algorithms the algorithm behaves well in precision,stability and engineering optimization.