BriefGPT.xyz
Jun, 2019
BatchBALD: 深度贝叶斯主动学习的高效多样化批量获取
BatchBALD: Efficient and Diverse Batch Acquisition for Deep Bayesian Active Learning
HTML
PDF
Andreas Kirsch, Joost van Amersfoort, Yarin Gal
TL;DR
BatchBALD是一种用于深度贝叶斯主动学习的可行近似方法，它计算了一批数据点与模型参数之间的互信息作为获取函数，并在获取批次中考虑依赖关系以提高数据效率并在标准基准测试中实现新的最先进性能.
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
→