Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2512.01067 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914297582452736 |
|---|---|
| author | Dramko, Evan Zhu, Yizhi Krivokapic, Aleksandar Hautier, Geoffroy Reps, Thomas Jermaine, Christopher Kyrillidis, Anastasios |
| author_facet | Dramko, Evan Zhu, Yizhi Krivokapic, Aleksandar Hautier, Geoffroy Reps, Thomas Jermaine, Christopher Kyrillidis, Anastasios |
| contents | Accurate structural relaxation is critical for advanced materials design. Traditional approaches built on physics-derived first-principles calculations are computationally expensive, motivating the creation of machine-learning interatomic potentials (MLIPs), which strive to faithfully reproduce first-principles computed forces. We propose a fine-tuning method to be used on a pretrained MLIP in which we create a fully-differentiable end-to-end simulation loop that optimizes the predicted final structures directly. Trajectories are unrolled and gradients are tracked through the entire relaxation. We show that this method consistently improves performance across all evaluated pretrained models; resulting in an average of roughly 32% reduction in prediction error. Interestingly, we show the process is robust to substantial variation in the relaxation setup, achieving negligibly different results across varied hyperparameter and procedural modifications. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_01067 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | On The Finetuning of MLIPs Through the Lens of Iterated Maps With BPTT Dramko, Evan Zhu, Yizhi Krivokapic, Aleksandar Hautier, Geoffroy Reps, Thomas Jermaine, Christopher Kyrillidis, Anastasios Materials Science Artificial Intelligence Machine Learning 68T07 I.2.1; J.2 Accurate structural relaxation is critical for advanced materials design. Traditional approaches built on physics-derived first-principles calculations are computationally expensive, motivating the creation of machine-learning interatomic potentials (MLIPs), which strive to faithfully reproduce first-principles computed forces. We propose a fine-tuning method to be used on a pretrained MLIP in which we create a fully-differentiable end-to-end simulation loop that optimizes the predicted final structures directly. Trajectories are unrolled and gradients are tracked through the entire relaxation. We show that this method consistently improves performance across all evaluated pretrained models; resulting in an average of roughly 32% reduction in prediction error. Interestingly, we show the process is robust to substantial variation in the relaxation setup, achieving negligibly different results across varied hyperparameter and procedural modifications. |
| title | On The Finetuning of MLIPs Through the Lens of Iterated Maps With BPTT |
| topic | Materials Science Artificial Intelligence Machine Learning 68T07 I.2.1; J.2 |
| url | https://arxiv.org/abs/2512.01067 |