Yahoo! Learning to Rank Challenge

6 papers18 directions
The Yahoo! Learning to Rank Challenge was a research venue active in 2011 that focused on the development and evaluation of machine learning algorithms for information retrieval. The venue centered on the "learning to rank" (LTR) problem, particularly within the context of web-search ranking. It provided a platform for researchers to present methodologies such as calibrated AdaBoost, extremely randomized trees, and initialized gradient boosted regression trees. An overview of the challenge and discussions on future directions in the field were also documented through its publications. Research presented at this venue addressed several core challenges in machine learning, including generalization through extreme randomization in tree ensembles and strategic model selection for efficient ensemble inference. A portion of the work focused on transfer learning, exploring its utility in low-data regimes, the use of smooth residual offsets, and methods to mitigate negative transfer via feature decoupling. Additionally, the venue featured research on robust domain adaptation and learning with asymmetric or corrupted label noise using latent transition matrices. The venue's corpus includes contributions from researchers such as Olivier Chapelle, Yi Chang, Balázs Kégl, and György Szarvas. The published papers detail methods for differentiable ranking and sorting for gradient-based learning and pointwise reranking for efficient text retrieval. By hosting specific tracks, such as the Transfer Learning Track, the venue facilitated the evaluation of YetiRank and other specialized ranking models, contributing to the documentation of ensemble-based and transfer-oriented learning techniques within the machine learning community.

Papers by Year

2011 6

Top Research Directions

Research directions most represented by papers from this venue.

Transfer Learning via Smooth Residual Offsets 1
Stabilizing Ranked Lists Against Data Perturbations 1
Surrogate Regret Bounds for Learning to Rank 1
Learning Theory for Pairwise Data Dependencies 1
The Limits of Ensemble Scaling 1
Feature Whitening in Representation Learning 1
Parallelization Constraints in Sequential Ensemble Training 1
Learning with Asymmetric and Class-Conditional Label Noise 1
Informed Initialization for Bayesian Optimization 1
Noise Transition Matrix Estimation for Noisy Labels 1
Memory Overhead in Multi-Model Inference Pipelines 1
Efficient Document Scoring for Real-Time Neural Ranking 1
Differentiable Relaxations for Sorting and Ranking 1
Ensembling Decision Trees for Complex Non-linear Modeling 1
Weighted Model Ensembling 1

Top Authors

AuthorPapers
Yi Chang2
Andrey Gulin1
Balázs Kégl1
Gilles Louppe1
György Szarvas1
Kilian Q Weinberger1
Pierre Geurts1
Robert Istvan Busa-Fekete1
Tie-Yan Liu1

Impactful Papers

Papers from this venue that shaped research clusters.

Foundational

Influential

Papers

TitleYear
Future directions in learning to rank2011
Learning to rank with extremely randomized trees2011
Ranking by calibrated AdaBoost2011
Web-Search Ranking with Initialized Gradient Boosted Regression Trees2011
Winning The Transfer Learning Track of Yahoo!'s Learning To Rank Challenge with YetiRank2011
Yahoo! Learning to Rank Challenge Overview2011