Enforcing Analytic Constraints in Neural Networks Emulating Physical Systems
Enforcing Analytic Constraints in Neural Networks Emulating Physical Systems
复制标题
DOI：
10.1103/physrevlett.126.098302
复制
发表时间：
2021-03-04
影响因子：
8.6
通讯作者：
Gentine, Pierre
中科院分区：
文献类型：
--
作者：
Beucler, Tom;Pritchard, Michael;Gentine, Pierre
神经网络能够高精度地模拟非线性物理系统，然而当违反基本约束时，它们可能会产生物理上不一致的结果。在此，我们介绍一种通过网络结构或损失函数中的约束在神经网络中强制实施非线性解析约束的系统方法。将其应用于气候建模的对流过程时，结构约束能在机器精度范围内强制满足守恒定律，且不会降低性能。强制实施约束还能减少受约束影响最大的输出子集的误差。
Neural networks can emulate nonlinear physical systems with high accuracy, yet they may produce physically inconsistent results when violating fundamental constraints. Here, we introduce a systematic way of enforcing nonlinear analytic constraints in neural networks via constraints in the architecture or the loss function. Applied to convective processes for climate modeling, architectural constraints enforce conservation laws to within machine precision without degrading performance. Enforcing constraints also reduces errors in the subsets of the outputs most impacted by the constraints.