This paper describes active learning algorithm used in AISTATS 2010 Active Learning Challenge as well as several of its extensions evaluated in the post-competition experiments. The algorithm consists of a pair of Regularized Parzen Window Classiers, one trained on full set of features and another on features ltered using Pearson correlation. Predictions of the two classiers are averaged to obtain the ensemble classier. Parzen Window classier was chosen because is an easy to implement lazy algorithm and has a single parameter, the kernel window size, that is determined by the cross-validation. The labeling schedule started by selecting random 20 examples and then continued by doubling the number of labeled examples in each round of active learning. A combination of random sampling and uncertainty sampling was used for querying. For the random sampling, examples were rst clustered using either all features or the ltered features (whichever resulted in higher cross-validated accuracy) and then the same number of random examples was selected from each cluster. Our algorithm ranked as the 5th overall, and was consistently ranked in the upper half of the competing algorithms. The challenge results show that Parzen Window classiers are less accurate than several competing learning algorithms used in the competition, but also indicate the success of the simple querying strategy that was employed. In the post-competition, we were able to improve the accuracy by using an ensemble of 5 Parzen Window classiers, each trained on features selected by dierent lters. We also explored how more involved querying during the initial stages of active learning and the pre-clustering querying strategy would inuence the performance of the proposed algorithm.
[1]
William A. Gale,et al.
A sequential algorithm for training text classifiers
,
1994,
SIGIR '94.
[2]
Arnold W. M. Smeulders,et al.
Active learning using pre-clustering
,
2004,
ICML.
[3]
H. Sebastian Seung,et al.
Query by committee
,
1992,
COLT '92.
[4]
David A. Cohn,et al.
Active Learning with Statistical Models
,
1996,
NIPS.
[5]
Zoran Obradovic,et al.
Feature Selection Filters Based on the Permutation Test
,
2004,
ECML.
[6]
Olivier Chapelle,et al.
Active Learning for Parzen Window Classifier
,
2005,
AISTATS.
[7]
Isabelle Guyon,et al.
Results of the Active Learning Challenge
,
2011,
Active Learning and Experimental Design @ AISTATS.