In this work, we propose a novel prior learning method for advancing generalization and uncertainty estimation in deep neural networks. The key idea is to exploit scalable and structured posteriors of neural networks as informative priors with generalization guarantees. Our learned priors provide expressive probabilistic representations at large scale, like Bayesian counterparts of pre-trained models on ImageNet, and further produce non-vacuous generalization bounds. We also extend this idea to a continual learning framework, where the favorable properties of our priors are desirable. Major enablers are our technical contributions: (1) the sums-of-Kronecker-product computations, and (2) the derivations and optimizations of tractable objectives that lead to improved generalization bounds. Empirically, we exhaustively show the effectiveness of this method for uncertainty estimation and generalization.

本文提出了一种基于先验学习的新方法，用于提高深度神经网络的泛化和不确定性估计，该方法利用可伸缩和结构化的神经网络后验作为具有泛化保证的信息先验。我们的学习先验在大规模上提供了具有表现力的概率表征，可以看作是在ImageNet上预训练模型的贝叶斯对应物，并进一步产生非平凡的泛化界限。我们还将这个想法扩展到了一个连续学习的框架中，其中我们的先验的有利特性是可取的。我们的技术贡献是（1）Kronecker积分和求和计算，以及（2）导出和优化可追踪的目标，从而导致改进的泛化边界。从实证上来说，我们详尽地展示了这种方法用于不确定性估计和泛化的有效性。

神经网络中学习表达先验的方法用于泛化和不确定性估计