Binary Trees for Classification, Regression, and Clustering, with Applications to Lossy Data Compression
Richard A. Olshen · 2005
This talk is a survey of binary tree-structured methods for classification, regression, survival analysis, and clustering. The discussion will include a survey of unifying themes, together with applications, and an introduction to mathematical issues that arise in studying their asymptotic properties. There will be special emphasis on the CAR/sup TM/ algorithms of Breiman et al., and on applications of the clustering algorithms to predictive, pruned, tree-structured vector quantization (predictive PTSVQ). The talk is a summary of collaborations with many authors over an eighteen year period.