Recent observations have advanced our understanding of the neural network
optimization landscape, revealing the existence of (1) paths of high accuracy
containing diverse solutions and (2) wider minima offering improved
performance. Previous methods observing diverse paths require mult
本研究提出了一种新颖的方案，通过构建神经网络参数的低维子空间（称为活跃子空间）来解决贝叶斯深度学习中由于参数空间的高维导致的计算复杂性限制。我们展示了显著降低的活跃子空间，通过 Monte Carlo（MC）采样方法或变分推断实现了可行和可扩展的贝叶斯推断，为各种回归任务提供了可靠的预测和鲁棒的不确定性估计。