Big Batch Bayesian Active Learning by Considering Predictive Probabilities

TL;DR

This paper introduces a new batch Bayesian active learning method that emphasizes predictive probabilities to better capture epistemic uncertainty, resulting in improved performance and efficiency over existing approaches.

Contribution

It proposes a novel acquisition function based on predictive probabilities that isolates epistemic uncertainty, enhancing batch size and computational speed.

Findings

Outperforms BatchBALD in accuracy and efficiency

Enables larger batch sizes due to faster evaluation

Better captures epistemic uncertainty in active learning

Abstract

We observe that BatchBALD, a popular acquisition function for batch Bayesian active learning for classification, can conflate epistemic and aleatoric uncertainty, leading to suboptimal performance. Motivated by this observation, we propose to focus on the predictive probabilities, which only exhibit epistemic uncertainty. The result is an acquisition function that not only performs better, but is also faster to evaluate, allowing for larger batches than before.