Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds
This work addresses the challenge of efficient data labeling in machine learning, offering a versatile solution for practical active learning problems, though it appears incremental as it builds on existing batch active learning methods.
The paper tackled the problem of batch active learning with deep neural networks by proposing the BADGE algorithm, which selects diverse and uncertain batches in a gradient space, resulting in consistent performance across various batch sizes and architectures without needing hand-tuned hyperparameters.
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.