Learning to rank using gradient descent

By C. Burges, Tal Shaked, Erin Renshaw, Ari Lazier, Matt Deeds, Nicole Hamilton, Gregory N. Hullender. 2005

3,228 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Retrieval & Search

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

Citations in the corpus

Dense Passage Retrieval for Open-Domain Question Answering (2020, background): “Similarly, in addition to negative loglikelihood, a popular option for ranking is triplet loss, which compares a positive passage and a negative one directly with respect to a question (Burges et al., 2005).”
Multi-Task Deep Neural Networks for Natural Language Understanding (2019, background): “The objective for the relevance ranking tasks follows the pairwise learning-to-rank paradigm (Burges et al., 2005; Huang et al., 2013).”
Representation Learning Using Multi-Task Deep Neural Networks for Semantic Classification and Information Retrieval (2015, methodology): “The objective for web search used in this paper follows the pair-wise learning-to-rank paradigm out-lined in (Burges et al., 2005).”
A Deep Relevance Matching Model for Ad-hoc Retrieval (2016, methodology): “LogCount-based Histogram (LCH): We apply logarithm over the count value in each bin, both to reduce the range, and to allow our model to more easily learn multiplicative relationships [1].”
ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT (2020, methodology): “;d := Õ i2jE q j max j2jE d j E i E T d j (3) ColBERT is di‡erentiable end-to-end. We train it to maximize the di‡erence in score between relevant and irrelevant documents via a pairwise RankNet loss [2]. We •ne-tune the BERT encoders and train from scratch the additional parameters (i.e., the convo”
Aligning AI With Shared Human Values (2020, methodology): “Models are tuned to output a scalar for each scenario while using the partial comparisons as the supervision signal (Burges et al., 2005).”
Counterfactual Estimation and Optimization of Click Metrics for Search Engines (2014, background): “Such an approach has been highly successful at facilitating easy comparison and improvement of ranking functions ( e.g. , [6, 32, 34]).”
Off-policy evaluation for slate recommendation (2016, background): “(cid:63) We deﬁne NDCG ( x, s ) := DCG ( x, s ) / DCG ( x ) where DCG (cid:63) ( x ) = max s ∈ S ( x ) DCG ( x, s ) , so NDCG takes values in [0 , 1] .”
Regularized Behavior Value Estimation (2021, background): “This ranking strategy can be formulated with a hinge-loss (Chen et al., 2009; Burges et al., 2005).”
BPR: Bayesian Personalized Ranking from Implicit Feedback (2009, methodology): “[1] optimize a neural network model for ranking using gradient descent.”
Pairwise Word Interaction Modeling with Deep Neural Networks for Semantic Similarity Measurement (2016, background): “Given two pieces of text, measuring their semantic textual similarity (STS) remains a fundamental problem in language research and lies at the core of many language processing tasks, including question answering (Lin, 2007), query ranking (Burges et al., 2005), and paraphrase generation (Xu, 2014).”
Multi-task learning for learning to rank in web search (2009, background): “RankNet [3] applies neural network and gradient descent to obtain a ranking function.”
An Introduction to Neural Information Retrieval (2018, neutral): “, 2007], neural networks [Burges et al., 2005], and boosted decision trees [Wu et al.”
Integer optimization methods for machine learning (2012, background): “There are a variety of other ranking algorithms that similarly minimize convex loss functions, such as RankProp and RankNet [Caruana et al., 1996, Burges et al., 2005].”
Hyperspherical Prototype Networks (2019, neutral): “ad of direct similarities. More formally, let W = fw 1;:::;w Kgdenote the word embeddings of the Kclasses. Using these embeddings, we deﬁne a loss function over all class triplets inspired by RankNet [5]: L PI = 1 jTj X (i;j;k)2T S ijklogS ijk (1 S ijk)log(1 S ijk); (6) where Tdenotes the set of all”
Conformer-Kernel with Query Term Independence for Document Retrieval (2020, methodology): “We train the model using the RankNet objective [Burges et al., 2005].”
WeBuildAI (2019, methodology): “This was based on the RankNet algorithm of [19].”
Document Comprehensiveness and User Preferences in Novelty Search Tasks (2015, background): “Early work on preference judgments in IR involved inferring preferences from those absolute judgments [2].”