DRC-ACM:一种精确的基于解析中心的分类器

DRC-ACM: Accurate Analytical Center Machine for Classification

ACM是一种基于版本空间 (versionspace)解析中心的分类器 ,它具有较好的泛化性能 但是由于冗余约束的存在 ,使得ACM分类器的解偏离主版本空间 (primeversionspace)解析中心 ,从而降低了分类器的泛化性能 ,同时冗余约束还将降低分类器的分类速度和存储效率 针对上述问题 ,提出了一种冗余约束增量约简算法 ,同时将增量约简算法与ACM分类器的算法结合起来 ,形成了一种去冗余约束的精确的ACM分类器 (DRC ACM ) 通过对Heart ,Thyroid ,Banana数据集的实验 ,证明DRC

Analytical center machine (ACM) has remarkable generalization performance, which is based on analytical center of version space From the analysis of geometry of machine learning and principle of ACM, it is shown that some constraints are redundant to the description of version space Redundant constraints push analytical center away from that of the prime version space so that the generalization performance degrades, and at the same time redundant constraints slow down the classifier and reduce the efficiency of storage due to non sparsity of ACM To overcome the above problems, an incremental algorithm is proposed to delete redundant constraints and embed into the frame of ACM that yields a non redundancy, accurate analytical center machine for classification called DRC ACM Experiments with Heart, Thyroid, Banana datasets demonstrate the validity of DRC ACM