We scale perceived distances of the core-set algorithm by a factor of uncertainty and search for low-confidence configurations, finding significant improvements in sample efficiency across CIFAR10/100 and SVHN image classification, especially in larger acquisition sizes. We show the necessity of our modifications and explain how the improvement is due to a probabilistic quadratic speed-up in the convergence of core-set loss, under assumptions about the relationship of model uncertainty and misclassification.
@article{arxiv.2202.04251,
title = {Improving greedy core-set configurations for active learning with uncertainty-scaled distances},
author = {Yuchen Li and Frank Rudzicz},
journal= {arXiv preprint arXiv:2202.04251},
year = {2022}
}