2016

Tighter bounds lead to improved classifiers

Roux, Nicolas Le

Understand

The standard approach to supervised classification involves the minimization of a log-loss as an upper bound to the classification error.

  • While this is a tight bound early on in the optimization, it overemphasizes the influence of incorrectly classified examples far from the decision boundary.
  • Updating the upper bound during the optimization leads to improved classification rates while transforming the learning into a sequence of minimization problems.
  • In addition, in the context where the classifier is part of a larger system, this modification makes it possible to link the performance of the classifier to that of the whole system, allowing the seamless introduction of external constraints.

Built on

  • Prodding the roc curve: Constrained optimization of classifier performance

    M. C. Mozer, R. H. Dodier, M. D. Colagrosso, C. Guerra-Salcedo, and R. H. Wolniewicz · 2001

    Earlier work this paper cites.

  • A parallel mixture of svms for very large scale problems

    R. Collobert, S. Bengio, and Y. Bengio · 2002

    Earlier work this paper cites.

  • The concave-convex procedure

    A. L. Yuille and A. Rangarajan · 2003

    Earlier work this paper cites.

  • Considering cost asymmetry in learning classifiers

    F. R. Bach, D. Heckerman, and E. Horvitz · 2006

    Earlier work this paper cites.

Similar

  • Trading convexity for scalability

    R. Collobert, F. Sinz, J. Weston, and L. Bottou · 2006

    Cited alongside, same era.

  • Robust support vector machine training via convex outlier ablation

    L. Xu, K. Crammer, and D. Schuurmans · 2006

    Cited alongside, same era.

  • Nonconvex online support vector machines

    c. Ertekin, L. Bottou, and C. L. Giles · 2011

    Cited alongside, same era.

  • Batch and online learning algorithms for nonconvex neyman-pearson classification

    G. Gasso, A. Pappaioannou, M. Spivak, and L. Bottou · 2011

    Cited alongside, same era.

  • Two high stakes challenges in machine learning

    L. Bottou

    Cited in the paper.

Then

  • A stochastic gradient method with an exponential convergence rate for finite training sets

    N. Le Roux, M. Schmidt, and F. Bach · 2012

    Later among the works it cites.

  • Counterfactual reasoning and learning systems: The example of computational advertising

    L. Bottou, J. Peters, J. Quinonero-Candela, D. X. Charles, D. M. Chickering, E. Portugaly, D. Ray, P. Simard, and E. Snelson · 2013

    Later among the works it cites.

  • Optimizing f-measures by cost-sensitive classification

    S. P. Parambath, N. Usunier, and Y. Grandvalet · 2014

    Later among the works it cites.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…