Statistical Analysis of Bayes Optimal Subset Ranking

By David Cossock, Tong Zhang. 2008

177 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Retrieval & Search, Optimization & Training

How researchers see this paper

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

In Artificial Intelligence

In Information Retrieval & Search

In Optimization & Training

Citations in the corpus

On Structured Prediction Theory with Calibrated Convex Surrogate Losses (2017, background): “However, most of the literature [e.g., 10, 13, 21, 1], focuses on calibration functions (in the form of regret bounds) for bipartite ranking, which is related to cost-sensitive binary classiﬁcation.”
"On the (Non-)existence of Convex, Calibrated Surrogate Losses for Ranking" (2012, background): “We describe in this section the formal framework of ranking we consider.”
Convex Calibration Dimension for Multiclass Loss Matrices (2014, background): “…in understanding consistency properties of surrogate risk minimization algorithms for various learning problems, such as binary clas-siﬁcation, multiclass 0-1 classiﬁcation, and various forms of ranking and multi-label prediction problems [18, 14, 34, 26, 2, 33, 28, 27, 8, 31, 10, 23, 4, 12, 15].”
Learning to Rank Using User Clicks and Visual Features for Image Retrieval (2015, methodology): “Treating learning to rank as either a regression or classification problem, Cossock and Zhang [4] formulated this problem using the objective function of a regression model, obtained through pointwise operations.”
On the Consistency of Ranking Algorithms (2010, methodology): “(We use these three losses to focus exclusively on learning in a pairwise setting— Cossock & Zhang (2008) learn using relevance scores, while Xia et al. (2008) require full ordered lists of re-sults as training data rather than pairs.)”
On the Consistency of Multi-Label Learning (2011, unknown): “…on optimizing a surrogate loss function has been well-studied for binary classiﬁcation (Zhang, 2004b; Steinwart, 2005; Bartlett et al., 2006), multi-class classiﬁcation (Zhang, 2004a; Tewari and Bartlett, 2007), learning to rank (Cossock and Zhang, 2008; Xia et al., 2008; Duchi et al., 2010), etc.”
Yahoo! Learning to Rank Challenge Overview (2010, methodology): “The objective function is of the form q,j (f (x q j ), l q j ) where can for instance be a regression loss (Cossock and Zhang, 2008) or a classification loss (Li et al., 2008).”
Learning Scoring Functions with Order-Preserving Losses and Standardized Supervision (2011, methodology): “Cossock & Zhang (2008) proved the consistency of a regression approach w.r.t. the DCG.”
Deep Multimodal Distance Metric Learning Using Click Constraints for Image Ranking (2017, methodology): “Specifically, Cossock and Zhang [42] formulated the problem of retrieval by using a regression model, which is trained through pointwise operations.”
An apple-to-apple comparison of Learning-to-rank algorithms in terms of Normalized Discounted Cumulative Gain (2012, background): “In subset ranking [3] (or web page ranking), the goal is to learn a ranking function that approximates the ideal partial ordering of a set of objects (or documents retrieved for the same query).”
Consistent Multiclass Algorithms for Complex Performance Measures (2015, background): “…2004a;b; Lee et al., 2004; Tewari & Bartlett, 2007), losses for specific problems such as multilabel classification (Gao & Zhou, 2011) and ranking (Cossock & Zhang, 2008; Xia et al., 2008; Duchi et al., 2010; Ravikumar et al., 2011; Buffoni et al., 2011; Calauzènes et al., 2012), and some work on…”
Margin-based Ranking and an Equivalence between AdaBoost and RankBoost (2009, background): “Several other recent works also consider this type of highly nonsymmetric setting for ranking (Dekel et al., 2004; Cossock and Zhang, 2008; Clemenc¸on and Vayatis, 2007; Shalev-Shwartz and Singer, 2006; Le and Smola, 2007).”
Convex Calibrated Surrogates for Low-Rank Loss Matrices with Applications to Subset Ranking Losses (2013, background): “As noted earlier, various loss functions are used in practice, and there has been much interest in understanding questions of consistency and calibration for these losses in recent years [9–15, 17].”
Surrogate regret bounds for bipartite ranking via strongly proper losses (2012, background): “Recently, there has been much interest in understanding statistical consistency and regret behavior of algorithms for a variety of ranking problems, including various forms of label/subset ranking as well as instance ranking problems [10, 8, 12, 3, 2, 30, 13, 21, 5, 9, 19, 29].”
Learning to Navigate for Fine-grained Classification (2018, methodology): “Generally speaking, these methods can be divided into three categories: the point-wise approach [9], pair-wise approach [18,4] and list-wise approach[6,44].”
A Theoretical Analysis of NDCG Type Ranking Measures (2013, methodology): “The research of consistency for ranking was initiated in [14, 17].”