Deep Batch Active Learning by Diverse, Uncertain Gradient Lower Bounds

By J. Ash, Chicheng Zhang, Akshay Krishnamurthy, J. Langford, Alekh Agarwal. 2019

1,024 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Mechanical Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Mechanical Engineering

Citations in the corpus

Fine-Tuning Language Models from Human Preferences (2019, background): “Examples of such techniques include selecting batches based on entropy considerations (Guo and Schuurmans, 2008), gradient-based metrics (Huang et al., 2016; Ash et al., 2019), or by attempting to distinguish labeled and unlabeled examples (Gissin and Shalev-Shwartz, 2019).”
Continual Deep Learning by Functional Regularisation of Memorable Past (2020, methodology): “For binary classification, our approach is equivalent to the “Confidence Sampling” approaches used in the Active Learning literature [4, 36], although in general it differs from them.”
On Warm-Starting Neural Network Training (2019, background): “Although this phenomenon has not received much direct attention from the research community, it seems to be common practice in deep active learning to retrain from scratch after every query step [6, 7]; popular deep active learning repositories on Github randomly reinitialize models after every sele”
RETRIEVE: Coreset Selection for Efficient and Robust Semi-Supervised Learning (2021, methodology): “To address this issue, we adopt a last-layer gradient approximation similar to [2, 39, 24, 23] by only considering the last classification layer gradients of the classifier model in RETRIEVE.”
GLISTER: Generalization based Data Subset Selection for Efficient and Robust Learning (2020, methodology): “Active Learning: Next, we compare GLISTER-ACTIVE to state-of-the-art active learning algorithms including FASS and BADGE.”
GRAD-MATCH: Gradient Matching based Data Subset Selection for Efficient Deep Model Training (2021, methodology): “To tackle this problem, we adopt a last-layer gradient approximation similar to (Ash et al., 2020; Mirzasoleiman et al., 2020a; Killamsetty et al., 2021) by only considering the last layer gradients for neural networks in GRADMATCH.”
A Survey of Deep Active Learning (2020, methodology): “In addition, previous heuristicbased AL [139] query strategies have proven to be ineffective when applied to DL [138]; therefore, for the one-by-one query strategy in classic AL, many researchers focus on the improvement of the batch sample query strategy [10, 51, 84, 183], taking both the amount of”