Semi-Supervised Neural Super-Resolution for Mesh-Based Simulations

2026-05-09

AI Summary

This work addresses the high computational cost of high-fidelity mesh simulation, which relies on fine-grained discretizations, and the data inefficiency of existing neural super-resolution methods that require abundant high-resolution supervision. To overcome these limitations, the authors propose SuperMeshNet, a semi-supervised framework that leverages a small set of paired low- and high-resolution meshes alongside a large corpus of unpaired low-resolution data. The approach employs two complementary message-passing neural networks (MPNNs) trained jointly with physics-informed inductive biases and a complementary learning mechanism. Remarkably, SuperMeshNet achieves significantly lower root mean square error (RMSE) than fully supervised baselines while using only 10% of the high-resolution training data, thereby substantially reducing data requirements without compromising reconstruction accuracy.

Abstract

Mesh-based simulations provide high-fidelity solutions to partial differential equations (PDEs), but achieving such accuracy typically requires fine meshes, leading to substantial computational overhead. Super-resolution techniques aim to mitigate this cost by reconstructing high-resolution (HR), high-fidelity solutions from low-cost, low-resolution (LR) counterparts. However, training neural networks for super-resolution often demands large amounts of expensive HR supervision data. To address this challenge, we propose SuperMeshNet, an HR data-efficient super-resolution framework for mesh-based simulations aided by message passing neural networks (MPNNs). At its core, SuperMeshNet introduces complementary learning, a semi-supervised approach that effectively leverages both 1) a small amount of paired LR-HR data and 2) abundant unpaired LR data via two jointly trained, complementary MPNN-based models. Additionally, our model is enriched by inductive biases, which are empirically shown to further improve super-resolution performance. Extensive experiments demonstrate that SuperMeshNet requires 90% less HR data to achieve even lower root mean square error (RMSE) than that of the fully supervised benchmark without the inductive biases. The source code and datasets are available at https://github.com/jykim-git/SuperMeshNet.git.

Problem

• super-resolution
• mesh-based simulations
• semi-supervised learning
• high-resolution data efficiency
• PDEs

Innovation

• semi-supervised learning
• neural super-resolution
• message passing neural networks
• complementary learning
• inductive biases