Diversity among base classifiers is known to be a necessary condition for improved performance of a classifier ensemble. However, there is an inevitable trade-off between accuracy and diversity, which is known as the accuracy/diversity dilemma. In this paper, accuracy and diversity are incorporated into a single measure, that is based on a spectral representation and computed between pairs of patterns of different class. Although the technique is only applicable to two-class problems, it is extended here to multi-class through Output Coding, and a comparison made between various weighted decoding schemes. © Springer-Verlag 2004.
CITATION STYLE
Windeatt, T. (2004). Spectral measure for multi-class problems. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 3077, 184–193. https://doi.org/10.1007/978-3-540-25966-4_18
Mendeley helps you to discover research relevant for your work.