Deep Learning for Physical Processes: Incorporating Prior Scientific\n Knowledge

2017/11/21 by Emmanuel de Bézenac, de Bezenac, Emmanuel, Arthur Pajot +3 · 14 citations
Computer Science · Earth and Planetary Sciences · Physics and Astronomy · #Artificial Intelligence (cs.AI) #Computational Physics and Python Applications #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Meteorological Phenomena and Simulations #Model Reduction and Neural Networks

paper · pdf · doi:10.48550/arxiv.1711.07970

openalex publication_date 2017/11/21 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/29

Abstract

We consider the use of Deep Learning methods for modeling complex phenomena\nlike those occurring in natural physical processes. With the large amount of\ndata gathered on these phenomena the data intensive paradigm could begin to\nchallenge more traditional approaches elaborated over the years in fields like\nmaths or physics. However, despite considerable successes in a variety of\napplication domains, the machine learning field is not yet ready to handle the\nlevel of complexity required by such problems. Using an example application,\nnamely Sea Surface Temperature Prediction, we show how general background\nknowledge gained from physics could be used as a guideline for designing\nefficient Deep Learning models. In order to motivate the approach and to assess\nits generality we demonstrate a formal link between the solution of a class of\ndifferential equations underlying a large family of physical phenomena and the\nproposed model. Experiments and comparison with series of baselines including a\nstate of the art numerical approach is then provided.\n

Citations

• A Recurrent Latent Variable Model for Sequential Data
• Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting
• Auto-Encoding Variational Bayes
• Accounting for an imperfect model in 4D‐Var
• Transformation-Based Models of Video Sequences
• Deep Residual Learning for Image Recognition

Cited by

• Efficient Generative Transformer Operators For Million-Point PDEs
• Adaptive Graph Rewiring to Mitigate Over-Squashing in Mesh-Based GNNs for Fluid Dynamics Simulations
• Parameter Estimation in Land Surface Models: Challenges and Opportunities With Data Assimilation and Machine Learning
• Video Prediction of Dynamic Physical Simulations With Pixel-Space Spatiotemporal Transformers
• Environment Inference for Learning Generalizable Dynamical System
• QDeepGR4J: Quantile-based ensemble of deep learning and GR4J hybrid rainfall-runoff models for extreme flow prediction with uncertainty quantification
• Harnessing artificial intelligence to fill global shortfalls in biodiversity knowledge
• Physics-informed machine learning: case studies for weather and climate modelling
• Analyzing Koopman approaches to physics-informed machine learning for long-term sea-surface temperature forecasting
• Can deep learning beat numerical weather prediction?
• InWaveSR: Topography-Aware Super-Resolution Network for Internal Solitary Waves
• A review of radar-based nowcasting of precipitation and applicable machine learning techniques
• Improving Long-term Autoregressive Spatiotemporal Predictions: A Proof of Concept with Fluid Dynamics
• Physics-aware nonparametric regression models for Earth data analysis

Related

• Physics-based Deep Learning
• Physics Informed Deep Learning (Part II): Data-driven Discovery of Nonlinear Partial Differential Equations
• Encoding physics to learn reaction-diffusion processes
• Physics Informed Deep Learning (Part I): Data-driven Solutions of Nonlinear Partial Differential Equations
• Physics-Guided Machine Learning for Scientific Discovery: An Application in Simulating Lake Temperature Profiles