Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds
Jordan T. Ash, Chicheng Zhang, Akshay Krishnamurthy, John Langford, Alekh Agarwal
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/JordanAsh/badgeOfficialpytorch★ 213
- github.com/webis-de/small-textpytorch★ 638
- github.com/decile-team/distilpytorch★ 155
- github.com/nlp-uoregon/famiepytorch★ 24
- github.com/duojun-huang/diana-cvpr2023pytorch★ 11
- github.com/dsba-lab/openalpytorch★ 10
Abstract
We design a new algorithm for batch active learning with deep neural network models. Our algorithm, Batch Active learning by Diverse Gradient Embeddings (BADGE), samples groups of points that are disparate and high-magnitude when represented in a hallucinated gradient space, a strategy designed to incorporate both predictive uncertainty and sample diversity into every selected batch. Crucially, BADGE trades off between diversity and uncertainty without requiring any hand-tuned hyperparameters. We show that while other approaches sometimes succeed for particular batch sizes or architectures, BADGE consistently performs as well or better, making it a versatile option for practical active learning problems.