Optimizing Classifier Performance via an Approximation to the Wilcoxon-Mann-Whitney Statistic

By Lian Yan, Robert H. Dodier, Michael C. Mozer, Richard H. Wolniewicz. International Conference on Machine Learning, 2003

305 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/df27dde10589455d290eeee6d0ae6ceeb83d0c6b

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Steven C. H. Hoi
2. Yoshua Bengio
3. Quoc V. Le
4. Dacheng Tao
5. Thorsten Joachims
6. Charu C. Aggarwal
7. Yi Yang
8. Sergey Levine

Citations in the corpus

• A support vector method for optimizing average precision (2007, methodology): “3.2 Finding the Most Violated Constraint Using OP 1 and optimizing to ROCArea loss (.roc ), the documents ranked before it, as opposed to j -1.”
• Supervised random walks: predicting and recommending links in social networks (2010, methodology): “Notice the strong cor­relation between AUC and WMW loss, i.e., as the value of the loss decreases, AUC increases.”
• A support vector method for multivariate performance measures (2005, methodology): “Also, methods for optimizing ROCArea have been proposed in the area of decision trees (Ferri et al., 2002), neural networks ( Yan et al., 2003; Herschtal & Raskutti, 2004), boosting (Cortes & Mohri, 2003; Freund et al., 1998), and SVMs (Herbrich et al., 2000; Rakotomamonjy, 2004).”
• Cost curves: An improved method for visualizing classifier performance (2006, methodology): “Recently, however, area under the ROC curve (AUC) has been used in some studies (Bradley, 1997; Karwath & King, 2002; Weiss and Provost, 2003; Yan et al., 2003).”
• AUC Maximization in the Era of Big Data and AI: A Survey (2022, methodology): “Second, models in diferent forms have been learned in the context of AUC maximization, including linear models [192], kernel models [68, 132], extreme learning machines [188], decision trees [45], neural networks [178], deep neural nets [194].”
• Learning to Rank with Nonsmooth Cost Functions (2006, methodology): “In fact for binary classification tasks, the pair-wise correct is the same as the AUC, which has led to work exploring optimizing the AUC using ranking algorithms [15, 3].”
• Online AUC Maximization (2011, methodology): “Two well-known algorithms are optimizing the Wilcoxon-Mann-Whitney statistic (Yan et al., 2003) and RankOpt (Herschtal & Raskutti, 2004) that adopts a differentiable approximation of AUC as its objective function.”