Comparing unsupervised learning methods for local structural identification in colloidal systems
A. Ulugöl, J.I. Bückmann, R. Yang, L.D. Hoitink, A. van Blaaderen, F. Smallenburg, L. Filion · 2025
Read Comparing unsupervised learning methods for local structural identification in colloidal systems by A. Ulugöl, J.I. Bückmann, R. Yang, L.D. Hoitink, A. van Blaaderen, F. Smallenburg, L. Filion (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-2509-07186.txt.
Description
Tiny particles that assemble themselves into crystals pose a sorting problem: how can researchers label each particle's local pattern? Alptuğ Ulugöl and 6 co-authors compare three unsupervised machine learning methods for that job in a 2025 research paper. The methods shrink complex particle data into simpler maps. They are principal component analysis, autoencoders, and UMAP, a popular nonlinear mapping tool.
The team first tests the methods on fluids and crystals of hard and charged spheres. They then try an icosahedral arrangement, a twenty-faced pattern, formed by spheres packed inside a round container, using both simulations and experiments. UMAP captures the complex structural features most reliably in every test. The comparison gives soft matter researchers a grounded pick when no one knows the right patterns in advance.
Catalog record
| CREATOR | A. Ulugöl, J.I. Bückmann, R. Yang, L.D. Hoitink, A. van Blaaderen, F. Smallenburg, L. Filion |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-07186