Data Dependence in Combining Classifiers

摘要: It has been accepted that multiple classifier systems provide a platform for not only performance improvement, but more efficient and robust pattern classification systems. A variety of combining methods have proposed in the literature some work focused on comparing categorizing these approaches. In this paper we present new categorization schemes based their dependence data patterns being classified. Combining can be totally independent from data, or they implicitly explicitly dependent data. is argued dependent, especially approaches represent highest potential improved performance. On basis categorization, an architecture explicit discussed. Experimental results to illustrate comparative according included.