Corpus ID: 2971050

Consistent optimization of AMS by logistic loss minimization

@article{Kotlowski2014ConsistentOO,
  title={Consistent optimization of AMS by logistic loss minimization},
  author={Wojciech Kotlowski},
  journal={ArXiv},
  year={2014},
  volume={abs/1412.2106}
}
  • Wojciech Kotlowski
  • Published 2014
  • Computer Science, Mathematics
  • ArXiv
  • In this paper, we theoretically justify an approach popular among participants of the Higgs Boson Machine Learning Challenge to optimize approximate median significance (AMS). The approach is based on the following two-stage procedure. First, a real-valued function is learned by minimizing a surrogate loss for binary classification, such as logistic loss, on the training sample. Then, a threshold is tuned on a separate validation sample, by direct optimization of AMS. We show that the regret of… CONTINUE READING
    How machine learning won the Higgs boson challenge
    2
    Higgs Boson Discovery with Boosted Trees
    69
    Consistent Classification Algorithms for Multi-class Non-Decomposable Performance Metrics

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 11 REFERENCES
    Convexity, Classification, and Risk Bounds
    966
    Weighted Classification Cascades for Optimizing Discovery Significance in the HiggsML Challenge
    7
    Surrogate regret bounds for bipartite ranking via strongly proper losses
    38
    Learning to discover: the Higgs boson machine learning challenge
    29
    Consistent Binary Classification with Generalized Performance Metrics
    81
    Convex Optimization
    34529
    Asymptotic formulae for likelihood-based tests of new physics
    2302
    The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd Edition
    12747