Learning functional components of PDEs from data using neural networks

2026-02-13 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Scientific Machine Learning · Depth: Advanced, quick

Summary

A new method embeds neural networks directly into partial differential equations (PDEs) to approximate unknown functional components that are difficult to measure directly. This approach extends existing scalar parameter-fitting workflows to recover entire functions from data, specifically demonstrated using nonlocal aggregation-diffusion equations. The neural networks are trained on steady-state data to approximate interaction kernels and external potentials. The research investigates how factors like the number of available solutions, their properties, sampling density, and measurement noise influence the success of function recovery. This technique allows the trained PDE to be used as a standard PDE for generating system predictions, offering a significant advantage for modeling complex systems.

Key takeaway

For AI Researchers and Research Scientists working with PDEs containing unmeasurable functional components, this method offers a robust way to recover those functions from data. You should consider embedding neural networks into your PDE models and training them on available steady-state data to enhance predictive capabilities, especially when direct measurement is infeasible.

Key insights

Neural networks can approximate unknown functional components within PDEs by embedding and training them with data.

Principles

• PDEs can incorporate neural networks as functional components.
• Training neural networks on data recovers unknown PDE functions.

Method

Embed neural networks into a PDE, then train them on steady-state data to approximate unknown functions like interaction kernels and external potentials, utilizing standard parameter-fitting workflows.

In practice

• Recover interaction kernels from aggregation-diffusion data.
• Approximate external potentials in complex PDE systems.

Topics

• Partial Differential Equations
• Neural Networks
• Function Recovery
• Scientific Machine Learning
• Nonlocal Aggregation-Diffusion

Best for: AI Researcher, AI Scientist, Research Scientist

Related on AIssential

• fTNN: a tensor neural network for fractional PDEs — fTNN is a deterministic tensor neural network method for fractional PDEs, using geometry-adapted integration and singularity-aware trial functions for high accuracy.
• Generalized Functional ANOVA in Closed-Form: A Unified View of Additive Explanations — A new method provides a tractable, explicit functional ANOVA decomposition for dependent continuous inputs, enhancing model interpretability.
• Functional Gradient Descent with Adaptive Representations — Adaptive representation in functional gradient descent ensures convergence to global minimizers by dynamically refining gradient approximations.
• Self-orthogonalizing attractor neural networks emerging from the free energy principle — Attractor networks emerge from the Free Energy Principle, yielding self-organizing, orthogonal representations and robust learning.
• Isotropic Fourier Neural Operators — Isotropic Fourier Neural Operators improve FNOs by enforcing spatial symmetries, boosting performance and reducing parameters.
• Neural Operator Processes for Probabilistic Operator Learning under Partial Observations — NOPs enable probabilistic operator learning from sparse data by unifying neural-process conditioning and neural-operator decoding.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.