HOPE: High-order Graph ODE For Modeling Interacting Dynamics
Jul 24, 2023
Speakers
Jingyang Yuan
Speaker · 0 followers
Zijie Huang
Speaker · 0 followers
Huiyu Jiang
Speaker · 0 followers
Yifang Qin
Speaker · 0 followers
Ming Zhang
Speaker · 1 follower
Yizhou Sun
Speaker · 0 followers
About
Leading graph ordinary differential equations (ODE) models have offered generalized strategies to extract useful information from massive interacting dynamical systems. They typically consist of a temporal graph encoder to provide the initial states and a neural ODE-based generative model to model the evolution of dynamical systems. However, these frameworks still have serious deficiencies in capacity and efficiency due to the failure to model high-order correlations embedded in long-term temporal trends and second-order mechanical laws. To tackle this, in this paper, we propose a novel model named High-Order graPh ODE (HOPE) for learning from dynamic interaction data. It first adopts a twin graph encoder to initialize the latent state representations of objects and edges, which consists of two branches to capture spatio-temporal correlations in complementary manners. More importantly, our HOPE utilizes a second-order graph ODE function which models the dynamics for both nodes and edges in the latent space respectively, which enables efficient learning of long-term dependencies from complex dynamical systems. Experiment results on a variety of datasets demonstrate both the effectiveness and efficiency of our proposed method.Leading graph ordinary differential equations (ODE) models have offered generalized strategies to extract useful information from massive interacting dynamical systems. They typically consist of a temporal graph encoder to provide the initial states and a neural ODE-based generative model to model the evolution of dynamical systems. However, these frameworks still have serious deficiencies in capacity and efficiency due to the failure to model high-order correlations embedded in long-term tempor…
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
BPipe: Memory-Balanced Pipeline Parallelism for Training Large Language Models
Watch later
Taebum Kim, …
Aiming for internal consistency, the 4th pillar of interactive learning
Watch later
Lowering the Pre-training Tax for Gradient-Based Subset Training
Watch later
Yeonju Ro, …
