基于蚁群聚类的码书设计

Codebook Design Based on Ant Colony Clustering

针对蚁群算法设计码书所存在的缺点,提出了改进,即引入了频率敏感方法,通过增加失真测度来减小蚂蚁重复选择同一个聚类的可能性,增加了选择的随机性,跳出了局部较小解,避免了停滞现象。有效地提高了其全局搜索能力。通过仿真实验,可以看出该算法码书的改善性能和寻优的高速率。

A new algorithm for codebook design, that combines ant colony with frequency sensitivity measure, is proposed by analyzing the shortcoming of codebook design based on ant colony clustering. The algorithm decrease the possibility of choosing the same clustering by increasing the distance between the centroid and vectors, in other word,it increases the randomicity in order to escape the local minimum and to prevent stagnation behaviour occurring. The experiment shows that we can effectively improve the ability to global search and the quality of codebook.