关键词: 频率指配, 模式分析, 核方法, 聚类分析, 核矩阵

Abstract: 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 kernel 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.

Key words: frequency assignment, pattern analysis, kernel method, cluster analysis, kernel matrix