Bayesian Neural Networks Uncertainty Quantification with Cubature Rules-猫眼课题宝

DOI：

10.1109/ijcnn48605.2020.9207214

发表时间：

2020-07

期刊：

2020 International Joint Conference on Neural Networks (IJCNN)

影响因子：

通讯作者：

中科院分区：

其他

文献类型：

作者：

Peng Wang;N. Bouaynaya;L. Mihaylova;Ji-kai Wang;Qibin Zhang;Renke He

贝叶斯神经网络通过考虑数据和网络模型中的随机性而成为强大的推理方法。神经网络输出的不确定性量化至关重要，特别是对于自动驾驶和危险天气预报等应用。然而，贝叶斯神经网络的理论分析方法仍然有限。本文对神经网络模型中的不确定性进行了数学量化，并提出了一种基于立方规则的计算效率高的不确定性量化方法，该方法可以捕获贝叶斯神经网络的逐层不确定性。所提出的方法近似的前两个时刻的后验分布的参数通过传播网络的非线性的体积点。仿真结果表明，该方法可以实现更多样化的分层不确定性量化的神经网络的结果与快速的收敛速度。

Bayesian neural networks are powerful inference methods by accounting for randomness in the data and the network model. Uncertainty quantification at the output of neural networks is critical, especially for applications such as autonomous driving and hazardous weather forecasting. However, approaches for theoretical analysis of Bayesian neural networks remain limited. This paper makes a step forward towards mathematical quantification of uncertainty in neural network models and proposes a cubature-rule-based computationally-efficient uncertainty quantification approach that captures layer-wise uncertainties of Bayesian neural networks. The proposed approach approximates the first two moments of the posterior distribution of the parameters by propagating cubature points across the network nonlinearities. Simulation results show that the proposed approach can achieve more diverse layer-wise uncertainty quantification results of neural networks with a fast convergence rate.

Bayesian Neural Networks Uncertainty Quantification with Cubature Rules

文献摘要

相似文献