Data-Driven Self-Supervised Learning for the Discovery of Solution Singularity for Partial Differential Equations
Difeng Cai, Paulina Sepúlveda · 2025
Read Data-Driven Self-Supervised Learning for the Discovery of Solution Singularity for Partial Differential Equations by Difeng Cai, Paulina Sepúlveda (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-23344.txt.
Description
Difeng Cai and Paulina Sepúlveda built a self-supervised method in twenty twenty-five for locating singularities hidden in data from partial differential equations. Singularities — spikes, jumps, thin layers — wreck numerical accuracy, and adaptive solvers need to know where they are. In the data-driven setting the input may be nothing but mesh vertices, with no labels at all.
Their framework learns from a filtering pretext task: two filters, one based on nearest neighbors and one on kernel density estimation, clean the raw points before learning. The authors show raw unfiltered data can produce pathological or wrong locations. Experiments covered interior circles, boundary layers, and concentric semicircles, and the method tolerated input noise and label corruption. Reliable singularity maps are the first step toward cheaper, adaptive simulation.
Catalog record
| CREATOR | Difeng Cai, Paulina Sepúlveda |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-23344