Agnostic active learning

We state and analyze the first active learning algorithm which works in the presence of arbitrary forms of noise. The algorithm, A2 (for Agnostic Active), relies only upon the assumption that the samples are drawn i.i.d. from a fixed distribution. A2 achieves an exponential improvement (i.e., requires only O ln 1 samples to find an -optimal classifier) over the usual sample complexity of supervised learning, for several settings considered before in the realizable case. These include learning threshold classifiers and learning homogeneous linear separators with respect to an input distribution which is uniform over the unit sphere. Key words: Active Learning, Agnostic Setting, Sample Complexity, Linear Separators, Uniform Distribution, Exponential Improvement.

• Type: Conference
• Year: 2006
• Where: ICML
• Authors: Maria-Florina Balcan, Alina Beygelzimer, John Langford
• Keywords: Active Learning Algorithm, Homogeneous Linear Separators, ICML 2006, Machine Learning, Usual Sample Complexity