BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning

Citations Over Time

Abstract

We develop BatchBALD, a tractable approximation to the mutual information between a batch of points and model parameters, which we use as an acquisition function to select multiple informative points jointly for the task of deep Bayesian active learning. BatchBALD is a greedy linear-time $1 - \frac{1}{e}$-approximate algorithm amenable to dynamic programming and efficient caching. We compare BatchBALD to the commonly used approach for batch data acquisition and find that the current approach acquires similar and redundant points, sometimes performing worse than randomly acquiring data. We finish by showing that, using BatchBALD to consider dependencies within an acquisition batch, we achieve new state of the art performance on standard benchmarks, providing substantial data efficiency improvements in batch acquisition.

Related Papers

• → Physician-Friendly Machine Learning: A Case Study with Cardiovascular Disease Risk Prediction(2019)71 cited
• → Artificial Intelligence, Machine Learning, and Medicine: A Little Background Goes a Long Way Toward Understanding(2021)29 cited
• → Application of Machine Learning in Animal Disease Analysis and Prediction(2020)26 cited
• → Sentiment Analysis by Using Supervised Machine Learning and Deep Learning Approaches(2020)3 cited
• → Breakdown of Machine Learning Algorithms(2022)1 cited