作者: I. Soraluze , C. Rodriguez , F. Boto , A. Cortes
DOI: 10.1007/978-3-540-24586-5_55
关键词: Random subspace method 、 NIST 、 Cascading classifiers 、 Image processing 、 Computer science 、 Algorithm 、 Speech processing 、 Classifier (linguistics) 、 Classifier (UML)
摘要: In this paper we present a way to reduce the computational cost of k-NN classifiers without losing classification power. Hierarchical or multistage have been built with purpose. These are designed putting incrementally trained into hierarchy and using rejection techniques in all levels apart from last. Results presented for different benchmark data sets: some standard sets taken UCI Repository Statlog Project, NIST Special Databases (digits upper-case lower-case letters). cases reduction is obtained maintaining recognition rate best individual classifier obtained.