neural-operator

Here I learn some stuff about neural operators. Starting with DeepONet and Fourier Neural Operators. FNOs not working very well yet.

OperatorForge · 世界级神经算子/科学机器学习系统（FNO / DeepONet / OpFuse），纯 NumPy 离线可跑、确定性可复现。作者：晨星

Data-free physics-informed neural operator for level-set interface advection. Spacetime FNO trained on the transport residual and eikonal constraint alone — no reference solutions. Supervised and hybrid baselines included. Two benchmarks: solid-body rotation and reversed single vortex.

A 13 module course on neural operators, from the convolution theorem to a physics foundation model. HTML lessons, notebooks, and three labs whose numbers all come from real runs.

This repository contains the document, presentation and scripts for inverse problems with neural operators.

A GLANCE into the Future: Grade-Guided Liquid AMD Network for Continuous-Time Evolution of AMD Progression from Fundus Images

HexaNTK — PyTorch training infrastructure for neural-operator runs on hexagonal grids (checkpoints, EMA, DDP guards, config/shape contracts for GINO/GAOT). Research code, no trained models yet.

Emergent symbolic reasoning based on VQ-tokenized neural operator rollouts, for use in dynamical cognitive memory, generalization, and reasoning across empirically grounded domains

This repository is the official implementation of the paper "RONOM: Reduced-Order Neural Operator Modeling" published in the SIAM Journal on Scientific Computing

GPU physics-informed neural operators for groundwater, benchmarked against per-well gray-box ODEs (PyTorch / PhysicsNeMo / CUDA)

verified differentiable NEGF solver for inverse design of quantum transport — benchmarks, gradient audits, FNO surrogate with OOD splits, and solver-confirmed inverse design

🚀 Build high-performance machine learning applications in pure Delphi with zero dependencies, fast native code, and advanced features for efficient computation.

RheOFormer: A transformer_based neural operator for simulation of complex fluids and flows. Covers 1D rheometric response predictions, and full 2D viscoelastic flow predictions on different geometries including irregular meshes.

Counterfactual exam shows a shipped FNO heat surrogate barely reacts to diffusivity changes while the solver does. Touchstone bench for PDE surrogates.

Foundation-model-inspired spatiotemporal neural operator for CFD-based offshore structural response prediction with synthetic CFD data, physics-calibrated residual inference, explainability, tests, and figures.

Research portfolio in Scientific AI: scientific machine learning, neural operators, digital twins, uncertainty quantification and sequential decision-making.

NeurIPS 2026 RealPDE Challenge — 1st in phase 1, Track 1 (Sim2Real), 4th Track 2. Real turbulent wake forecasting with calibrated uncertainty bounds.

Do physics-informed neural networks extrapolate better? Pre-registered benchmark of PINNs vs data-driven MLP and Fourier neural operators for nonlinear ultrashort-pulse propagation (NLSE) outside the training range.

Neural operators mapping QFT actions to boundary correlators in Euclidean AdS2, with live WebGPU surfaces and browser-side inference

Notebook-first Fourier Neural Operator study for real NOAA OISST forecasting