Ranking Measures and Loss Functions in Learning to Rank

By Tie-Yan Liu, Yanyan Lan, Zhiming Ma, Hang Li. 2009

241 citations recorded by the source corpus.

How researchers see this paper

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

Citations in the corpus

Regularized Behavior Value Estimation (2021, background): “This ranking strategy can be formulated with a hinge-loss (Chen et al., 2009; Burges et al., 2005).”
Leveraging Unlabeled Data for Crowd Counting by Learning to Rank (2018, background): “There are several works which have studied how to learn to rank, and they focus on learning a ranking function from ground-truth rankings [7, 27].”
Decoupled Contrastive Learning (2021, background): “JRPC is more similar to the ranking loss (Chen et al., 2009), which collects and pushes away the positive pairs and negative pairs.”
Autoregressive Reasoning over Chains of Facts with Transformers (2020, background): “This loss is shown by Chen et al. (2009) to maximize a lower bound on the MAP.”
A Neural Autoencoder Approach for Document Ranking and Query Refinement in Pharmacogenomic Information Retrieval (2018, methodology): “We use a pairwise hinge loss (Chen et al., 2009) and compare relevant documents with irrelevant documents.”
TF-Ranking: Scalable TensorFlow Library for Learning-to-Rank (2018, methodology): “oss in our library, the pairwise logistic loss is defined as: ℓˆ( y, ˆ)= Õn j=1 Õn k=1 I(yj > yk)log (1+exp −j))) (9) where I(·)is the indicator function. Finally, as an example of a listwise loss [11], our library provides the implementation of Softmax Cross-Entropy, ListNet [7], and ListMLE [43”
An Introduction to Neural Information Retrieval (2018, unknown): “Pairwise loss generally has the following form [Chen et al., 2009],”
TransFG: A Transformer Architecture for Fine-grained Recognition (2021, methodology): “A typical strategy is to use these local features for classiﬁcation individually and adopt a rank loss (Chen et al. 2009) to maintain consistency between the quality of bounding boxes and their ﬁnal probability output.”
LARA: Attribute-to-feature Adversarial Learning for New-item Recommendation (2020, methodology): “To thoroughly measure our model and the baselines, we employ precision (P@K), mean average precision (M@K), and normalized discounted cumulative gain (NDCG@K) as the evaluation metrics [6] to measure the item cold-start recommendation performance from different angles.”
RankIQA: Learning from Rankings for No-Reference Image Quality Assessment (2017, background): “These approaches learn a ranking function from ground-truth rankings by minimizing a ranking loss [3].”
A Deep Look into Neural Ranking Models for Information Retrieval (2019, methodology): “In contrast to pointwise methods where the final ranking loss is the sum of loss on each document, pairwise loss functions are computed based on the permutations of all possible document pairs [96].”
IRGAN: A Minimax Game for Unifying Generative and Discriminative Information Retrieval Models (2017, methodology): “In addition to the logistic function log ( 1 + exp (− z )) , it is possible to make use of other pairwise ranking loss functions [7], such as the hinge function ( 1 − z ) + (as used in Ranking SVM [16]) and the exponential function exp (− z ) (as used in RankBoost [11]), to define the probability…”
Learning to Rank for Information Retrieval and Natural Language Processing (2011, unknown): “The pairwise loss functions in Ranking SVM, RankBoost, and RankNet are upper bounds of (1-NDCG) [20].”
Tuning as Ranking (2011, background): “We cast tuning as a ranking problem (Chen et al., 2009), where the explicit goal is to learn to correctly rank candidate translations.”