Optimism in Active Learning with Gaussian Processes

In the context of Active Learning for classification, the classification error depends on the joint distribution of samples and their labels which is initially unknown. Online estimation of this distribution, for the purpose of minimizing the error, involves a trade-off between exploration and exploitation. This is a common problem in machine learning for which multi-armed bandit theory, building upon the paradigm of Optimism in the Face of Uncertainty, has been proven very efficient. We introduce two novel algorithms that use Optimism in the Face of Uncertainty along with Gaussian Processes for the Active Learning problem. Evaluations lead on real-world datasets show that these new algorithms compare positively to state-of-the-art methods.