论文部分内容阅读
为了有效地控制仿射传播聚类的全局搜索和局部搜索,将仿射传播聚类视作一个搜索能量函数最小值的方法,提出一种基于动态阻尼因子策略的动态阻尼仿射传播聚类.使用两阶段动态阻尼因子策略,使得仿射传播聚类迭代过程中阻尼因子不再是静态固定值,而是自适应的动态改变值,构造了1种线性策略和2种非线性策略来动态改变阻尼因子的值.在标准数据集上的实验结果表明:动态阻尼仿射传播聚类避免了仿射传播聚类发生震荡,同时可以加快仿射传播聚类的收敛.
In order to control the global search and local search of affine propagation clustering effectively, affine propagation clustering is considered as a method to search the minimum of energy function. A dynamic damping affine propagation clustering based on dynamic damping factor strategy is proposed. Using two-stage dynamic damping factor strategy, the damping factor in the iterative process of affine propagation clustering is no longer a static fixed value, but an adaptive dynamic change value. A linear strategy and two non-linear strategies are constructed to dynamically change The experimental results on a standard dataset show that the dynamic damping affine propagation clustering avoids the convolution of affine propagation clustering and accelerates the convergence of affine propagation clustering.