UPAL: Unbiased Pool Based Active Learning

Ravi Ganti, Alexander Gray
Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics, PMLR 22:422-431, 2012.

Abstract

In this paper we address the problem of pool based active learning, and provide an algorithm, called UPAL, that works by minimizing the unbiased estimator of the risk of a hypothesis in a given hypothesis space. For the space of linear classifiers and the squared loss we show that UPAL is equivalent to an exponentially weighted average forecaster. Exploiting some recent results regarding the spectra of random matrices allows us to analyze UPAL with squared losses for the noiseless setting. Empirical comparison with an active learner implementation in Vowpal Wabbit, and a previously proposed pool based active learner implementation show good empirical performance and better scalability.

Cite this Paper

BibTeX
@InProceedings{pmlr-v22-ganti12, title = {UPAL: Unbiased Pool Based Active Learning}, author = {Ganti, Ravi and Gray, Alexander}, booktitle = {Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics}, pages = {422--431}, year = {2012}, editor = {Lawrence, Neil D. and Girolami, Mark}, volume = {22}, series = {Proceedings of Machine Learning Research}, address = {La Palma, Canary Islands}, month = {21--23 Apr}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v22/ganti12/ganti12.pdf}, url = {https://proceedings.mlr.press/v22/ganti12.html}, abstract = {In this paper we address the problem of pool based active learning, and provide an algorithm, called UPAL, that works by minimizing the unbiased estimator of the risk of a hypothesis in a given hypothesis space. For the space of linear classifiers and the squared loss we show that UPAL is equivalent to an exponentially weighted average forecaster. Exploiting some recent results regarding the spectra of random matrices allows us to analyze UPAL with squared losses for the noiseless setting. Empirical comparison with an active learner implementation in Vowpal Wabbit, and a previously proposed pool based active learner implementation show good empirical performance and better scalability.} }
Endnote
%0 Conference Paper %T UPAL: Unbiased Pool Based Active Learning %A Ravi Ganti %A Alexander Gray %B Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2012 %E Neil D. Lawrence %E Mark Girolami %F pmlr-v22-ganti12 %I PMLR %P 422--431 %U https://proceedings.mlr.press/v22/ganti12.html %V 22 %X In this paper we address the problem of pool based active learning, and provide an algorithm, called UPAL, that works by minimizing the unbiased estimator of the risk of a hypothesis in a given hypothesis space. For the space of linear classifiers and the squared loss we show that UPAL is equivalent to an exponentially weighted average forecaster. Exploiting some recent results regarding the spectra of random matrices allows us to analyze UPAL with squared losses for the noiseless setting. Empirical comparison with an active learner implementation in Vowpal Wabbit, and a previously proposed pool based active learner implementation show good empirical performance and better scalability.
RIS
TY - CPAPER TI - UPAL: Unbiased Pool Based Active Learning AU - Ravi Ganti AU - Alexander Gray BT - Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics DA - 2012/03/21 ED - Neil D. Lawrence ED - Mark Girolami ID - pmlr-v22-ganti12 PB - PMLR DP - Proceedings of Machine Learning Research VL - 22 SP - 422 EP - 431 L1 - http://proceedings.mlr.press/v22/ganti12/ganti12.pdf UR - https://proceedings.mlr.press/v22/ganti12.html AB - In this paper we address the problem of pool based active learning, and provide an algorithm, called UPAL, that works by minimizing the unbiased estimator of the risk of a hypothesis in a given hypothesis space. For the space of linear classifiers and the squared loss we show that UPAL is equivalent to an exponentially weighted average forecaster. Exploiting some recent results regarding the spectra of random matrices allows us to analyze UPAL with squared losses for the noiseless setting. Empirical comparison with an active learner implementation in Vowpal Wabbit, and a previously proposed pool based active learner implementation show good empirical performance and better scalability. ER -
APA
Ganti, R. & Gray, A.. (2012). UPAL: Unbiased Pool Based Active Learning. Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 22:422-431 Available from https://proceedings.mlr.press/v22/ganti12.html.

Related Material