交互动力系统的深度状态空间模型的廉价和确定性推理（英文版）

英文标题：Cheap and Deterministic Inference for Deep State-Space Models of Interacting Dynamical Systems中文摘要：本文提出了一种基于图神经网络的深度状态空间模型，用于建模交互式动态系统，且预测分布为多峰高斯混合模型，同时提出了结构化的协方差近似方法，可以适用于具有多个智能体的系统，并在自动驾驶数据集上进行了测试并与现有技术进行了比较。英文摘要：Graph neural networks are often used to model interacting dynamical systemssince they gracefully scale to systems with a varying and high number ofagents. While there has been much progress made for deterministic interactingsystems, modeling is much more challenging for stochast