The field of physics-informed neural networks and operator learning is rapidly advancing, with a focus on improving accuracy, efficiency, and interpretability. Recent developments have introduced new frameworks, such as FNODE, Neptune, and HyPINO, which have demonstrated superior performance in various applications, including data-driven simulation, parameter estimation, and solving inverse problems. These advancements have the potential to transform various fields, including engineering, healthcare, and physics. Noteworthy papers include FNODE, which learns acceleration vector fields directly from trajectory data, and Neptune, which infers parameter fields from sparse measurements. Additionally, HyPINO has shown strong zero-shot accuracy on benchmark problems, outperforming existing methods.
Advances in Physics-Informed Neural Networks and Operator Learning
Sources
FNODE: Flow-Matching for data-driven simulation of constrained multibody systems
Estimating Parameter Fields in Multi-Physics PDEs from Scarce Measurements
Theory Foundation of Physics-Enhanced Residual Learning
An Evolutionary Multi-objective Optimization for Replica-Exchange-based Physics-informed Operator Learning Network
Disentangling Slow and Fast Temporal Dynamics in Degradation Inference with Hierarchical Differential Models
Accelerating PDE Solvers with Equation-Recast Neural Operator Preconditioning
Efficient Transformer-Inspired Variants of Physics-Informed Deep Operator Networks
Non-Asymptotic Performance Analysis of DOA Estimation Based on Real-Valued Root-MUSIC
Computational Fluid Dynamics Optimization of F1 Front Wing using Physics Informed Neural Networks
Fisher information flow in artificial neural networks
CLINN: Conservation Law Informed Neural Network for Approximating Discontinuous Solutions
Graph neural networks for learning liquid simulations in dynamic scenes containing kinematic objects
Solving Inverse Acoustic Obstacle Scattering Problem with Phaseless Far-Field Measurement Using Deep Neural Network Surrogates
HyPINO: Multi-Physics Neural Operators via HyperPINNs and the Method of Manufactured Solutions
SPINN: An Optimal Self-Supervised Physics-Informed Neural Network Framework
Multi-Stage Graph Neural Networks for Data-Driven Prediction of Natural Convection in Enclosed Cavities
Data-Efficient Time-Dependent PDE Surrogates: Graph Neural Simulators vs Neural Operators
Information-Theoretic Bounds and Task-Centric Learning Complexity for Real-World Dynamic Nonlinear Systems
DEQuify your force field: More efficient simulations using deep equilibrium models