Differentiable Physics-informed Graph Networks

Abstract

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.

11 pages, 9 figures

References in corpus (9)

• Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
• Inductive Representation Learning on Large Graphs
• Hidden Physics Models: Machine Learning of Nonlinear Partial Differential Equations
• Physics Informed Deep Learning (Part I): Data-driven Solutions of Nonlinear Partial Differential Equations
• A simple neural network module for relational reasoning
• Graph networks as learnable physics engines for inference and control
• A Compositional Object-Based Approach to Learning Physical Dynamics
• Visual Interaction Networks
• Hodge Laplacians on graphs

Cited by in corpus (17)

• Learning to schedule job-shop problems: Representation and policy learning using graph neural network and reinforcement learning
• Hamiltonian Graph Networks with ODE Integrators
• Latent Space Data Assimilation by using Deep Learning
• Physics-guided Design and Learning of Neural Networks for Predicting Drag Force on Particle Suspensions in Moving Fluids
• Inductive Graph Neural Networks for Spatiotemporal Kriging
• Learning continuous-time PDEs from sparse data with graph neural networks
• STAN: Spatio-Temporal Attention Network for Pandemic Prediction Using Real World Evidence
• Neural Weighted A*: Learning Graph Costs and Heuristics with Differentiable Anytime A*
• A Deep Learning Approach for Predicting Spatiotemporal Dynamics From Sparsely Observed Data
• Climate Modeling with Neural Diffusion Equations
• Uncertainty Quantification and Software Risk Analysis for Digital Twins in the Nearly Autonomous Management and Control Systems: A Review
• Neural Pharmacodynamic State Space Modeling
• Accurately Solving Physical Systems with Graph Learning
• Deep learning reveals hidden interactions in complex systems
• Physics-aware Spatiotemporal Modules with Auxiliary Tasks for Meta-Learning
• Model discovery in the sparse sampling regime
• Physics-Coupled Spatio-Temporal Active Learning for Dynamical Systems