频率指配问题求解的模式分析核方法  

Using kernel methods for pattern analysis to solve frequency assignment problem

作　　者：杨化斌[1] 林中[1] 孙俊[1] 

机构地区：[1]空军工程大学电讯工程学院

出　　处：《计算机工程与应用》2013年第21期168-171,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61174162;No.61101102);航空科学基金(No.20100796004)

摘　　要：为有效解决频率指配问题,提出了一种解决该问题的模式分析核方法,算法利用频率指配方案的评估函数值构建核矩阵,以核矩阵为基础在特征空间中运行聚类分析算法,对频率指配方案相似性进行度量,完成频率指配方案的归类划分。优化结果可直接作为跳频指配结果,也可作为其他优化算法的初始解。该算法在电台数量规模较大的应用中体现出良好的性能,算法结果作为蚁群、遗传算法的初始解,后继算法收敛速度明显提高。The kernel methods for pattern analysis are applied to solving the frequency assignment problem. The algorithm takes randomly assigned solutions and their corresponding values of object function to construct kemel matrix. Then based on the kernel matrix, this paper uses cluster analysis algorithm in the eigen space to measure similarity of solutions and classify solutions. Optimized results of the cluster analysis algorithm can be directly applied to actual engineering project, and also can be used as the initial input of other optimization algorithms. Using optimized results of the cluster analysis algorithm as the initial solution, algorithms like ant colony algorithm and genetic algorithm reflect high efficiency in the application that has large-scale radio stations. Convergence rate is improved significantly.

关 键 词：频率指配 模式分析 核方法 聚类分析 核矩阵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]