Bayesian Active Learning by Disagreements: A Geometric Perspective
This work addresses the challenge of enhancing active learning efficiency for machine learning practitioners by providing a novel geometric approach, though it appears incremental as it builds upon existing BALD methods.
The paper tackles the problem of improving Bayesian active learning by disagreements (BALD) through a geometric perspective, introducing GBALD which constructs core-sets on ellipsoids instead of spheres to prevent low-representative elements and reduce redundant estimations, resulting in outperforming existing methods like BALD and BatchBALD in experiments.
We present geometric Bayesian active learning by disagreements (GBALD), a framework that performs BALD on its core-set construction interacting with model uncertainty estimation. Technically, GBALD constructs core-set on ellipsoid, not typical sphere, preventing low-representative elements from spherical boundaries. The improvements are twofold: 1) relieve uninformative prior and 2) reduce redundant estimations. Theoretically, geodesic search with ellipsoid can derive tighter lower bound on error and easier to achieve zero error than with sphere. Experiments show that GBALD has slight perturbations to noisy and repeated samples, and outperforms BALD, BatchBALD and other existing deep active learning approaches.