论文标题
基于图形层次结构的神经网络
Equivariant Graph Hierarchy-Based Neural Networks
论文作者
论文摘要
模棱两可的图形神经网络（EGN）在表征多体物理系统的动力学方面具有强大的功能。现有的EGN进行平坦的消息传递，但该消息无法捕获复杂系统的空间/动力层次结构，尤其是限制了子结构发现和全局信息融合。在本文中，我们提出了基于层次层次结构的图形网络（EGHNS），该图由三个关键组成部分组成：广义eproivariant矩阵消息传递（EMMP），e-pool和e-uppool。特别是，EMMP能够提高常规eproivariast消息传递的表达性，电子池将低级节点的数量分配到高级簇中，而E-uppool则利用高级信息来更新低级节点的动力学。正如他们的名字所暗示的那样，E-Pool和E-uppool都可以符合物理对称性。大量的实验评估验证了我们的EGHN对多种应用的有效性，包括多对象动力学模拟，运动捕获和蛋白质动力学建模。
Equivariant Graph neural Networks (EGNs) are powerful in characterizing the dynamics of multi-body physical systems. Existing EGNs conduct flat message passing, which, yet, is unable to capture the spatial/dynamical hierarchy for complex systems particularly, limiting substructure discovery and global information fusion. In this paper, we propose Equivariant Hierarchy-based Graph Networks (EGHNs) which consist of the three key components: generalized Equivariant Matrix Message Passing (EMMP) , E-Pool and E-UpPool. In particular, EMMP is able to improve the expressivity of conventional equivariant message passing, E-Pool assigns the quantities of the low-level nodes into high-level clusters, while E-UpPool leverages the high-level information to update the dynamics of the low-level nodes. As their names imply, both E-Pool and E-UpPool are guaranteed to be equivariant to meet physic symmetry. Considerable experimental evaluations verify the effectiveness of our EGHN on several applications including multi-object dynamics simulation, motion capture, and protein dynamics modeling.
