Data-Efficient and Fast Machine Learning Molecular Dynamics through Integrated Active Learning and Knowledge Distillation
Xiliang Lian, Alfredo Pasquarello · 2026
Read Data-Efficient and Fast Machine Learning Molecular Dynamics through Integrated Active Learning and Knowledge Distillation by Xiliang Lian, Alfredo Pasquarello (2026) 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/pmc-13564361.txt.
Description
Xiliang Lian and Alfredo Pasquarello speed up machine learning models of atomic forces in work dated two thousand twenty-six. Such models drive fast simulations of how molecules move, but training them eats data and time. Two model families were compared on liquid water, a standard test case.
One family needed about three times less training data but ran ten times slower. Starting from a pretrained base cut data needs far more. The team then distilled the slow model's skill into the fast family through a cheap active learning loop. The distilled model took about ten times less compute to build and matched the others on water's structure, shaking, and spread.
Catalog record
| CREATOR | Xiliang Lian, Alfredo Pasquarello |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |