Winning The Transfer Learning Track of Yahoo!’s Learning To Rank Challenge with YetiRank

Andrey Gulin, Igor Kuralenok, Dimitry Pavlov
Proceedings of the Learning to Rank Challenge, PMLR 14:63-76, 2011.

Abstract

The problem of ranking the documents according to their relevance to a given query is a hot topic in information retrieval. Most learning-to-rank methods are supervised and use human editor judgements for learning. In this paper, we introduce novel pairwise method called YetiRank that modifies Friedman’s gradient boosting method in part of gradient computation for optimization and takes uncertainty in human judgements into account. Proposed enhancements allowed YetiRank to outperform many state-of-the-art learning to rank methods in offline experiments as well as take the first place in the second track of the Yahoo! learning-to-rank contest. Even more remarkably, the first result in the learning to rank competition that consisted of a transfer learning task was achieved without ever relying on the bigger data from the “transfer-from” domain.

Cite this Paper


BibTeX
@InProceedings{pmlr-v14-gulin11a, title = {Winning The Transfer Learning Track of Yahoo!’s Learning To Rank Challenge with YetiRank}, author = {Gulin, Andrey and Kuralenok, Igor and Pavlov, Dimitry}, booktitle = {Proceedings of the Learning to Rank Challenge}, pages = {63--76}, year = {2011}, editor = {Chapelle, Olivier and Chang, Yi and Liu, Tie-Yan}, volume = {14}, series = {Proceedings of Machine Learning Research}, address = {Haifa, Israel}, month = {25 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v14/gulin11a/gulin11a.pdf}, url = {https://proceedings.mlr.press/v14/gulin11a.html}, abstract = {The problem of ranking the documents according to their relevance to a given query is a hot topic in information retrieval. Most learning-to-rank methods are supervised and use human editor judgements for learning. In this paper, we introduce novel pairwise method called YetiRank that modifies Friedman’s gradient boosting method in part of gradient computation for optimization and takes uncertainty in human judgements into account. Proposed enhancements allowed YetiRank to outperform many state-of-the-art learning to rank methods in offline experiments as well as take the first place in the second track of the Yahoo! learning-to-rank contest. Even more remarkably, the first result in the learning to rank competition that consisted of a transfer learning task was achieved without ever relying on the bigger data from the “transfer-from” domain.} }
Endnote
%0 Conference Paper %T Winning The Transfer Learning Track of Yahoo!’s Learning To Rank Challenge with YetiRank %A Andrey Gulin %A Igor Kuralenok %A Dimitry Pavlov %B Proceedings of the Learning to Rank Challenge %C Proceedings of Machine Learning Research %D 2011 %E Olivier Chapelle %E Yi Chang %E Tie-Yan Liu %F pmlr-v14-gulin11a %I PMLR %P 63--76 %U https://proceedings.mlr.press/v14/gulin11a.html %V 14 %X The problem of ranking the documents according to their relevance to a given query is a hot topic in information retrieval. Most learning-to-rank methods are supervised and use human editor judgements for learning. In this paper, we introduce novel pairwise method called YetiRank that modifies Friedman’s gradient boosting method in part of gradient computation for optimization and takes uncertainty in human judgements into account. Proposed enhancements allowed YetiRank to outperform many state-of-the-art learning to rank methods in offline experiments as well as take the first place in the second track of the Yahoo! learning-to-rank contest. Even more remarkably, the first result in the learning to rank competition that consisted of a transfer learning task was achieved without ever relying on the bigger data from the “transfer-from” domain.
RIS
TY - CPAPER TI - Winning The Transfer Learning Track of Yahoo!’s Learning To Rank Challenge with YetiRank AU - Andrey Gulin AU - Igor Kuralenok AU - Dimitry Pavlov BT - Proceedings of the Learning to Rank Challenge DA - 2011/01/26 ED - Olivier Chapelle ED - Yi Chang ED - Tie-Yan Liu ID - pmlr-v14-gulin11a PB - PMLR DP - Proceedings of Machine Learning Research VL - 14 SP - 63 EP - 76 L1 - http://proceedings.mlr.press/v14/gulin11a/gulin11a.pdf UR - https://proceedings.mlr.press/v14/gulin11a.html AB - The problem of ranking the documents according to their relevance to a given query is a hot topic in information retrieval. Most learning-to-rank methods are supervised and use human editor judgements for learning. In this paper, we introduce novel pairwise method called YetiRank that modifies Friedman’s gradient boosting method in part of gradient computation for optimization and takes uncertainty in human judgements into account. Proposed enhancements allowed YetiRank to outperform many state-of-the-art learning to rank methods in offline experiments as well as take the first place in the second track of the Yahoo! learning-to-rank contest. Even more remarkably, the first result in the learning to rank competition that consisted of a transfer learning task was achieved without ever relying on the bigger data from the “transfer-from” domain. ER -
APA
Gulin, A., Kuralenok, I. & Pavlov, D.. (2011). Winning The Transfer Learning Track of Yahoo!’s Learning To Rank Challenge with YetiRank. Proceedings of the Learning to Rank Challenge, in Proceedings of Machine Learning Research 14:63-76 Available from https://proceedings.mlr.press/v14/gulin11a.html.

Related Material