论文信息 - BORN AGAIN TREES

BORN AGAIN TREES

Tree predictors such as CART or C4.5 are often not as accurate as neural nets or use of multiple trees. But these latter methods lead to predictors whose structure is difficult to understand, whereas trees have a universal simplicity. Because of this, it is appealing to try and find tree representations of more complex predictors. We study tree representers of multiple tree predictors. These representers are larger, more stable and more accurate than trees grown the usual way. For this reason, we call them "born again" trees.

L. Breiman | Nong Shang

[1] Leo Breiman,et al. Classification and Regression Trees , 1984 .

[2] Jude W. Shavlik,et al. in Advances in Neural Information Processing , 1996 .

[3] Thomas G. Dietterich,et al. Error-Correcting Output Coding Corrects Bias and Variance , 1995, ICML.

[4] Corinna Cortes,et al. Boosting Decision Trees , 1995, NIPS.

[5] Yoav Freund,et al. A decision-theoretic generalization of on-line learning and an application to boosting , 1995, EuroCOLT.

[6] Yoav Freund,et al. Experiments with a New Boosting Algorithm , 1996, ICML.

[7] J. Ross Quinlan,et al. Bagging, Boosting, and C4.5 , 1996, AAAI/IAAI, Vol. 1.

[8] Leo Breiman,et al. Bias, Variance , And Arcing Classifiers , 1996 .