BriefGPT.xyz
Oct, 2022
贝叶斯深度学习的加速线性化拉普拉斯近似
Accelerated Linearized Laplace Approximation for Bayesian Deep Learning
HTML
PDF
Zhijie Deng, Feng Zhou, Jun Zhu
TL;DR
通过开发一种Nystrom近似方法来加速线性化变种(Laplace Approximation)和神经切向核(NTKs)之间的联系, 以解决Bayesian神经网络中的非常规低效率问题。该方法通过自动区分前向模式来实现,具有可靠的理论保证,并在规模性和性能方面表现出许多优点。
Abstract
laplace approximation
(LA) and its
linearized variant
(LLA) enable effortless adaptation of pretrained deep neural networks to
bayesian neural ne
→