← Back to all news
arXiv AI By Georg Maierhofer, Miles Cranmer, Payel Mukhopadhyay

Mesh-Native Tokenisation: Retrofitting Foundation Models to Unstructured Data

Read the original on arXiv AI →

Mesh‑Native Tokenisation introduces Mesh‑Native Convolution (MNC), a locality‑aware tokenisation layer that encodes unstructured mesh‑based data similarly to convolutional neural networks. This allows physics foundation models trained on grid data to be retrofitted to arbitrary unstructured meshes with minimal retraining, achieving competitive accuracy even in zero‑shot deployment. The paper provides theoretical validation of MNC’s consistency with classical CNN encoders and demonstrates its superior performance over existing baselines after only a few fine‑tuning epochs.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

Related stories

arXiv AI
Oct 5

Geometry Meets Physics: Data-Efficient Pre-Training for Unstructured Neural PDE Solvers

The paper presents a data‑efficient pre‑training framework for neural surrogate models solving PDEs on unstructured 3D geometries. It introduces a geometry‑driven strategy for steady‑state problems that uses intrinsic shape descriptors, and a physics‑driven approach for transient problems that generates synthetic PDE data online. Experiments show faster convergence, higher data efficiency, and improved accuracy during fine‑tuning, especially in low‑data settings.

By Luis Medrano-Navarro, Giacomo Baldan, Qiang Liu, Benjamin Holzschuh, Jan Hagnberger, Mathias Niepert, Nils Thuerey
fine-tuning
More like this →
arXiv AI
Aug 19

ChannelFlow-Tools: A Configuration-Driven Pipeline for Generating Machine-Learning-Ready Datasets of 3D Obstructed Channel Flows

ChannelFlow-Tools is an open‑source, configuration‑driven pipeline that generates machine‑learning‑ready datasets for three‑dimensional obstructed channel flows. It combines procedural obstacle geometry generation across six shape families, signed‑distance‑field voxelisation, lattice‑Boltzmann simulation, and packaging into ML‑ready tensors, all driven by reproducible configuration files. The pipeline is validated through mesh‑integrity audits, SDF representation checks, solver benchmarks, and data‑integrity audits, and it has been used to train surrogate models (3D U‑Net, FNO, U‑FNO) that learn geometry‑to‑flow mappings and exhibit physically interpretable behaviour on out‑of‑distribution splits.

By Shubham Kavane, Lukas Schr\"oder, Kajol Kulkarni, Fernando Gonzalez, Harald Koestler
benchmarks
More like this →
arXiv Machine Learning
Jul 10

PGD-NO: A Neural Operator with Precomputed Geometry Decomposition for 3D Million-scale Physics Simulations

arXiv:2607. 08025v1 Announce Type: new Abstract: While neural PDE solvers have demonstrated significant potential for accelerating engineering simulations, existing architectures remain constrained by high memory consumption and the single node bottleneck, where the maximum processable mesh resolution is strictly limited by the VRAM of a single compute unit.

By Weiheng Zhong, Jing Bi, Victor Oancea, Hadi Meidani
benchmarkssafety
More like this →