BriefGPT.xyz
Apr, 2023
将未标记的数据融入贝叶斯神经网络中
Incorporating Unlabelled Data into Bayesian Neural Networks
HTML
PDF
Mrinank Sharma, Tom Rainforth, Yee Whye Teh, Vincent Fortuin
TL;DR
使用对比学习框架，提出一种适用于未标记数据的贝叶斯神经网络（BNNs）的先验分布，得到自监督学习的标签效率和贝叶斯方法的规范不确定性估计，最后证明了该方法在半监督学习和低预算主动学习问题中的数据效率优势。
Abstract
We develop a
contrastive framework
for learning better prior distributions for
bayesian neural networks
(BNNs) using unlabelled data. With this framework, we propose a practical BNN algorithm that offers the labe
→