Stumping along a Summary for Exploration & Exploitation Challenge 2011

Christophe Salperwyck, Tanguy Urvoy
Proceedings of the Workshop on On-line Trading of Exploration and Exploitation 2, PMLR 26:86-97, 2012.

Abstract

The \emphPascal Exploration & Exploitation challenge 2011 seeks to evaluate algorithms for the online website content selection problem. This article presents the solution we used to achieve second place in this challenge and some side-experiments we performed. The methods we evaluated are all structured in three layers. The first layer provides an online summary of the data stream for continuous and nominal data. Continuous data are handled using an online quantile summary. Nominal data are summarized with a hash-based counting structure. With these techniques, we managed to build an accurate stream summary with a small memory footprint. The second layer uses the summary to build predictors. We exploited several kinds of trees from simple decision stumps to deep multivariate ones. For the last layer, we explored several combination strategies: online bagging, exponential weighting, linear ranker, and simple averaging.

Cite this Paper

BibTeX
@InProceedings{pmlr-v26-salperwyck12a, title = {Stumping along a Summary for Exploration & Exploitation Challenge 2011}, author = {Salperwyck, Christophe and Urvoy, Tanguy}, booktitle = {Proceedings of the Workshop on On-line Trading of Exploration and Exploitation 2}, pages = {86--97}, year = {2012}, editor = {Glowacka, Dorota and Dorard, Louis and Shawe-Taylor, John}, volume = {26}, series = {Proceedings of Machine Learning Research}, address = {Bellevue, Washington, USA}, month = {02 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v26/salperwyck12a/salperwyck12a.pdf}, url = {https://proceedings.mlr.press/v26/salperwyck12a.html}, abstract = {The \emphPascal Exploration & Exploitation challenge 2011 seeks to evaluate algorithms for the online website content selection problem. This article presents the solution we used to achieve second place in this challenge and some side-experiments we performed. The methods we evaluated are all structured in three layers. The first layer provides an online summary of the data stream for continuous and nominal data. Continuous data are handled using an online quantile summary. Nominal data are summarized with a hash-based counting structure. With these techniques, we managed to build an accurate stream summary with a small memory footprint. The second layer uses the summary to build predictors. We exploited several kinds of trees from simple decision stumps to deep multivariate ones. For the last layer, we explored several combination strategies: online bagging, exponential weighting, linear ranker, and simple averaging.} }
Endnote
%0 Conference Paper %T Stumping along a Summary for Exploration & Exploitation Challenge 2011 %A Christophe Salperwyck %A Tanguy Urvoy %B Proceedings of the Workshop on On-line Trading of Exploration and Exploitation 2 %C Proceedings of Machine Learning Research %D 2012 %E Dorota Glowacka %E Louis Dorard %E John Shawe-Taylor %F pmlr-v26-salperwyck12a %I PMLR %P 86--97 %U https://proceedings.mlr.press/v26/salperwyck12a.html %V 26 %X The \emphPascal Exploration & Exploitation challenge 2011 seeks to evaluate algorithms for the online website content selection problem. This article presents the solution we used to achieve second place in this challenge and some side-experiments we performed. The methods we evaluated are all structured in three layers. The first layer provides an online summary of the data stream for continuous and nominal data. Continuous data are handled using an online quantile summary. Nominal data are summarized with a hash-based counting structure. With these techniques, we managed to build an accurate stream summary with a small memory footprint. The second layer uses the summary to build predictors. We exploited several kinds of trees from simple decision stumps to deep multivariate ones. For the last layer, we explored several combination strategies: online bagging, exponential weighting, linear ranker, and simple averaging.
RIS
TY - CPAPER TI - Stumping along a Summary for Exploration & Exploitation Challenge 2011 AU - Christophe Salperwyck AU - Tanguy Urvoy BT - Proceedings of the Workshop on On-line Trading of Exploration and Exploitation 2 DA - 2012/05/02 ED - Dorota Glowacka ED - Louis Dorard ED - John Shawe-Taylor ID - pmlr-v26-salperwyck12a PB - PMLR DP - Proceedings of Machine Learning Research VL - 26 SP - 86 EP - 97 L1 - http://proceedings.mlr.press/v26/salperwyck12a/salperwyck12a.pdf UR - https://proceedings.mlr.press/v26/salperwyck12a.html AB - The \emphPascal Exploration & Exploitation challenge 2011 seeks to evaluate algorithms for the online website content selection problem. This article presents the solution we used to achieve second place in this challenge and some side-experiments we performed. The methods we evaluated are all structured in three layers. The first layer provides an online summary of the data stream for continuous and nominal data. Continuous data are handled using an online quantile summary. Nominal data are summarized with a hash-based counting structure. With these techniques, we managed to build an accurate stream summary with a small memory footprint. The second layer uses the summary to build predictors. We exploited several kinds of trees from simple decision stumps to deep multivariate ones. For the last layer, we explored several combination strategies: online bagging, exponential weighting, linear ranker, and simple averaging. ER -
APA
Salperwyck, C. & Urvoy, T.. (2012). Stumping along a Summary for Exploration & Exploitation Challenge 2011. Proceedings of the Workshop on On-line Trading of Exploration and Exploitation 2, in Proceedings of Machine Learning Research 26:86-97 Available from https://proceedings.mlr.press/v26/salperwyck12a.html.

Related Material