Research On Human Learning Optimization Algorithms And Solving The 0-1 Knapsack Problem

Swarm intelligence optimization algorithm is inspired by group cooperation behaviors in various animal societies,which usually make it more effective in the optimization process towards the best solution.All individuals of the solution space must work together and cooperate closely to find a solution which is good enough,and this will be the candidate solution for the next round of optimization process.Human learning optimization algorithm is some swarm intelligence computation algorithm based on learning processes of human beings,the algorithm seeks the global optimal solution through the coordination of different learning operators.It takes advantage of the characteristics of human skills on learning new knowledge,and it is a new and very potential optimization method which can often produce better optimization results than many traditional intelligence optimization algorithms.This new algorithm has some common features with other traditional swarm intelligence algorithms,but it also has problems in convergence rate and accuracy,and it may be unstable and easy to fall into local optimum.Aiming at overcome the shortcomings of basic human learning optimization algorithm,in this thesis a new concept of "pairing mechanism" is proposed for the first time.Firstly,according to the operator cooperation mechanism in the basic human learning optimization,"paired learning" is introduced after individual learning to the algorithm and mutual-study will be encouraged,and then better solutions will be recommended to the social learning process for re-learning.In addition,the main parameters of the algorithm which may affect the results of calculation are adjusted,and conduct many simulation experiments,which can help avoid problems of poor efficiency and/or low precision caused by improper parameter settings.Finally,MATLAB simulation results of 10 bench test functions are presented graphically to show the superiority of human learning optimization algorithm based on pairing mechanism,compared with the traditional algorithm.To further verify the capability of human learning optimization algorithm with pairing mechanism in solving practical problems,the 0-1 knapsack problem is chosen as the application background.Firstly,the calculation process and algorithm implementation of 0-1 knapsack problem are discussed in detail,then we use different swarm intelligence algorithms to carry out simulation experiments in MATLAB,and with each algorithm 10 independent experiments are conducted respectively.Finally,all simulation results are compared with each other and show that,in most cases,our human learning optimization algorithm with pairing mechanism performs usually better than the original human learning optimization algorithm,simulated annealing algorithm,ant colony algorithm,etc.