Empirical Margin Distributions and Bounding the Generalization Error of Combined Classifiers

@inproceedings{Koltchinskii2000EmpiricalMD,
  title={Empirical Margin Distributions and Bounding the Generalization Error of Combined Classifiers},
  author={Vladimir Koltchinskii and Dmitry Panchenko},
  year={2000}
}
We prove new probabilistic upper bounds on generalization error of complex classi ers that are combinations of simple classi ers. Such combinations could be implemented by neural networks or by voting methods of combining the classi ers, such as boosting and bagging. The bounds are in terms of the empirical distribution of the margin of the combined classi er. They are based on the methods of the theory of Gaussian and empirical processes (comparison inequalities, symmetrization method… CONTINUE READING
Highly Influential
This paper has highly influenced 30 other papers. REVIEW HIGHLY INFLUENTIAL CITATIONS

Citations

Publications citing this paper.
Showing 1-10 of 250 extracted citations

A manifold framework of multiple-kernel learning for hyperspectral image classification

Neural Computing and Applications • 2016
View 5 Excerpts
Highly Influenced

Some topics on similarity metric learning

View 12 Excerpts
Highly Influenced

References

Publications referenced by this paper.
Showing 1-10 of 29 references

Risk Bounds for Model Selection via Penalization

A. Barron, L. Birg e, P. Massart
Probability Theory and Related Fields, • 1999
View 4 Excerpts
Highly Influenced

A probabilistic theory of pattern recognition

L. Devroye, L. Gy or, G. Lugosi
1996
View 5 Excerpts
Highly Influenced

Weak convergence and Empirical Processes

A. van der Vaart, J. Wellner
With Applications to Statistics. Springer-Verlag, • 1996
View 5 Excerpts
Highly Influenced

Rademacher penalties and structural risk minimization

IEEE Trans. Information Theory • 2001
View 1 Excerpt

About the constants in Talagrand ’ s concentration inequalities for empirical processes

P. Massart
Ann . Probab . • 2000

Similar Papers

Loading similar papers…