Improvement And Application Of Invasive Weed Optimization Algorithm

In the age of big data and artificial intelligence,the rapid development of computer science and technology has completely changed people's traditional way of life.In order to provide a better user experience,it becomes more urgent to solve problems in engineering practices,such as multimedia classification,target retrieval,data mining and so on.Such problems can often be transformed into high-dimensional,nonlinear function optimization problems.The best solution is reached by obtaining the optimal solution of objective function.For these problems are nondifferentiable and their real-time requirements of the system are high,the traditional optimization algorithms are difficult to apply;therefore,to search for more efficient optimization algorithms has become the key to solving such problems.The emergence of intelligent optimization algorithms provides practical solutions for such problems.The study of intelligent optimization algorithms is of great significance in engineering practice.This thesis conducts in-depth research on Invasive Weed Optimization(IWO)in intelligent optimization algorithms,aiming at the phenomenon that it easily falls into a local in highdimensional problems,and applies the fractional blobal best formation(FGBF)strategy to improve it.Aimed at the problem that the IWO does not have the adaptability in the optimization of variable dimension of the decision variables,the thesis advances the multidimensional by combining the Particle Swarm Optimization.In order to further improve the convergence of the multidimensional IWO and reduce its dependence on parameter settings,the FGBF strategy and tropism growth strategy are used to modify the multidimensional IWO.At last,the improved multidimensional IWO is applied to nonlinear function optimization and image segmentation.The experimental results show that in the nonlinear function optimization problem,the improved IWO has less dependence on the parameter setting and it converges faster,and in the unsupervised image segmentation problem,the optimum solution and the mean solution that the improved multidimensional IWO reaches are smaller than the particle swarm algorithm does.