Elements of information theory
Elements of information theory
Rigorous learning curve bounds from statistical mechanics
Machine Learning - Special issue on COLT '94
Self bounding learning algorithms
COLT' 98 Proceedings of the eleventh annual conference on Computational learning theory
COLT '99 Proceedings of the twelfth annual conference on Computational learning theory
Microchoice bounds and self bounding learning algorithms
COLT '99 Proceedings of the twelfth annual conference on Computational learning theory
Expected Error Analysis for Model Selection
ICML '99 Proceedings of the Sixteenth International Conference on Machine Learning
On the Convergence Rate of Good-Turing Estimators
COLT '00 Proceedings of the Thirteenth Annual Conference on Computational Learning Theory
Generalization Bounds for Decision Trees
COLT '00 Proceedings of the Thirteenth Annual Conference on Computational Learning Theory
Localized Rademacher Complexities
COLT '02 Proceedings of the 15th Annual Conference on Computational Learning Theory
Quantitatively tight sample complexity bounds
Quantitatively tight sample complexity bounds
Improving SVM classifiers training using artificial samples
ICCOMP'07 Proceedings of the 11th WSEAS International Conference on Computers
Combinatorial shell bounds for generalization ability
Pattern Recognition and Image Analysis
Human Action Segmentation and Recognition Using Discriminative Semi-Markov Models
International Journal of Computer Vision
| Hi-index | 0.00 |
Haussler, Kearns, Seung and Tishby introduced the notion of a shell decomposition of the union bound as a means of understanding certain empirical phenomena in learning curves such as phase transitions. Here we use a variant of their ideas to derive an upper bound on the generalization error of a hypothesis computable from its training error and the histogram of training errors for the hypotheses in the class. In most cases this new bound is significantly tighter than traditional bounds computed from the training error and the cardinality of the class. Our results can also be viewed as providing a rigorous foundation for a model selection algorithm proposed by Scheffer and Joachims.