Specifying Weight Priors in Bayesian Deep Neural Networks with Empirical Bayes
This addresses a bottleneck in Bayesian deep learning for researchers and practitioners needing scalable uncertainty estimation, though it is incremental as it builds on existing variational inference methods.
The paper tackles the challenge of specifying meaningful weight priors in Bayesian deep neural networks for scalable variational inference, proposing the MOPED method which uses empirical Bayes to set informed priors and demonstrates reliable uncertainty quantification across tasks like image classification and diabetic retinopathy diagnosis.
Stochastic variational inference for Bayesian deep neural network (DNN) requires specifying priors and approximate posterior distributions over neural network weights. Specifying meaningful weight priors is a challenging problem, particularly for scaling variational inference to deeper architectures involving high dimensional weight space. We propose MOdel Priors with Empirical Bayes using DNN (MOPED) method to choose informed weight priors in Bayesian neural networks. We formulate a two-stage hierarchical modeling, first find the maximum likelihood estimates of weights with DNN, and then set the weight priors using empirical Bayes approach to infer the posterior with variational inference. We empirically evaluate the proposed approach on real-world tasks including image classification, video activity recognition and audio classification with varying complex neural network architectures. We also evaluate our proposed approach on diabetic retinopathy diagnosis task and benchmark with the state-of-the-art Bayesian deep learning techniques. We demonstrate MOPED method enables scalable variational inference and provides reliable uncertainty quantification.