摘要
本文提出了一种基于非均匀变异的演化算法,该算法具有搜索步长依概率减小,开始阶段均匀搜索全部空间,结束阶段仅在当前解周田搜索的自适应特性。用其求解三个方程组,一个是线性方程组,一个是摄像机四点定位(P4P)问题的高效求解,一个是“油层结垢”的非线性方程组,并与经典的BFGS方法和混合符号数值方法进行了对比,结果显示BFGS方法在求解践性方程组时显出优势。对四点摄像机定位问题,所做前期工作很少,但效果和混合符号数值方法差不多。对非线性方程组,演化算法则显出很好的优势和鲁棒性,说明BFGS方法对解决具有良好性质的线性问题非常有效,而演化方法对解决复杂非线性问题有很强的适用性。
An evolutionary algorithm based on the non-uniform mutation is proposed in this paper whose search step size is decreasing in probability with the run of the algorithm.This algorithm is able to search the total space in the former stage and only searches its neighborhood in the later stage.Three equations are solved with this algorithm that are linear equation group,the geometry of camera pose from four points(P4P)and the deposited oil layers problems.Our algorithm compare with the BFGS and a hybrid symbolic and numerical computing approach.BFGS shows advantage for the linear equation.The results of o^ar EA method and the hybrid approach are comparable,however,many specific/heuristic knowledge are used by their method for the P4P problem.The EA method shows great dominance to the BFGS for the real-world nonlinear equation.It proved that BFGS is suitable to the linear problems with excellent properties and EA method has predominance on the complex nonlinear questions.
作者
赵新超
Xinchac Zhao(School of Sciences.Beijing University of Posts and Telecommunications,Beijing 100876,China)
出处
《通讯和计算机(中英文版)》
2005年第8期20-26,共7页
Journal of Communication and Computer
基金
本文得到2004年中国科学院研究生科学与社会实践资助专项(创新研究类):“演化计算在蛋白质折叠中的应用”资助.
关键词
演化算法
BFGS
P4P问题
方程组
最优化
Evolutionary Algorithm
P4P Problem
Equation Solving
Global Optimization
