Variational Linearized Laplace Approximation for Bayesian Deep Learning

The Linearized Laplace Approximation (LLA) has been recently used to perform uncertainty estimation on the predictions of pre-trained deep neural networks (DNNs). However, its widespread application is hindered by significant computational costs, particularly in scenarios with a large number of training points or DNN parameters. Consequently, additional approximations of LLA, such as Kronecker-factored or diagonal approximate GGN matrices, are utilized, potentially compromising the model's performance. To address these challenges, we propose a new method for approximating LLA using a variational …

application approximation bayesian bayesian deep learning computational costs cs.lg deep learning dnn laplace approximation networks neural networks parameters predictions stat.ml training uncertainty