Differentiable Physics-informed Graph Networks

By Sungyong Seo, Yan Liu. 2019

70 citations recorded by the source corpus.

While physics conveys knowledge of nature built from an interplay between observations and theory, it has been considered less importantly in deep neural networks. Especially, there are few works leveraging physics behaviors when the knowledge is given less explicitly. In this work, we propose a novel architecture called Differentiable Physics-informed Graph Networks (DPGN) to incorporate implicit physics knowledge which is given from domain experts by informing it in latent space. Using the concept of DPGN, we demonstrate that climate prediction tasks are significantly improved. Besides the experiment results, we validate the effectiveness of the proposed module and provide further applications of DPGN, such as inductive learning and multistep predictions.

Subfields: Artificial Intelligence, Graph Learning, AI for Physical Sciences

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Chuizheng Meng
2. Sungyong Seo
3. Nicolas Thomé
4. Vincent Le Guen
5. Yan Liu
6. Zixuan Ke
7. Bing Liu

In Graph Learning

1. Sungyong Seo
2. Yan Liu

In AI for Physical Sciences

1. Sungyong Seo
2. Nicolas Thomé
3. Vincent Le Guen
4. Yan Liu

Citations in the corpus

• Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges (2021, neutral): “2019) and graph neural network (Seo and Liu, 2019) backbones.”
• Physics-aware Difference Graph Networks for Sparsely-Observed Dynamics (2020, background): “Seo & Liu (2019) proposed a physics-informed regularizer to impose data-specific physics equations.”
• Disentangling Physical Dynamics From Unknown Factors for Unsupervised Video Prediction (2020, background): “Based on the connection between numerical schemes for solving PDEs ( e.g. Euler, Runge-Kutta) and residual neural networks [71, 37, 8, 78], several specific architectures were designed for predicting and identifying dynamical systems [15, 35, 47].”
• Learning to schedule job-shop problems: representation and policy learning using graph neural network and reinforcement learning (2021, background): “As a result, GNN models can successfully perform their tasks without additional training in various fields including robot control, wind power estimation, and city-scale weather prediction.[32, 48, 39].”