Title :
Greedy growing of tree-structured classification rules using a composite splitting criterion
Author :
Nobel, Andrew B.
Author_Institution :
Dept. of Stat., North Carolina Univ., Chapel Hill, NC, USA
Abstract :
We establish the Bayes risk consistency of an unsupervised greedy-growing algorithm that produces tree-structured classifiers from labeled training vectors. The algorithm employs a composite splitting criterion equal to a weighted sum of Bayes risk and Euclidean distortion
Keywords :
Bayes methods; image coding; statistical analysis; trees (mathematics); vector quantisation; Bayes risk consistency; Euclidean distortion; composite splitting criterion; image compression; labeled training vectors; tree-structured classification rules; unsupervised greedy-growing algorithm; weighted sum; Algorithm design and analysis; Binary trees; Classification tree analysis; Euclidean distance; Image coding; Information theory; Quantization; Random variables; Statistics; Voting;
Conference_Titel :
Information Theory and Statistics, 1994. Proceedings., 1994 IEEE-IMS Workshop on
Conference_Location :
Alexandria, VA
Print_ISBN :
0-7803-2761-6
DOI :
10.1109/WITS.1994.513860
