Rethinking the role of frames for SE(3)-invariant crystal structure modeling
Keywords: Materials Science, Invariant Networks, Transformer, Physics-Informed ML
TL;DR: Propose dynamic atom-wise frames aligned with interatomic interactions on the structure rather than with the structure itself.
Abstract: Crystal structure modeling using geometric graph neural networks is important in various machine learning applications in materials science. In these applications, capturing SE(3)-invariant geometric features in crystal structures is a fundamental requirement for these networks. One approach is to model with orientation-standardized structures through structure-aligned coordinate systems called `frames.' However, unlike molecules, determining frames for crystal structures is not trivial due to their infinite and highly symmetric nature. In the search for effective frames for crystals, we point out that existing work assumes a statically fixed frame for each structure based solely on its structural information, regardless of the task under consideration. Here, we rethink the role of frames, *questioning whether such simplistic alignment with the structure is sufficient*, and propose the concept of *dynamic frames*. While accommodating the infinite and symmetric nature of crystals, these frames give each atom its own dynamic view of the structure, focusing only on those atoms actively interacting with it. We demonstrate this concept by utilizing the attention mechanism in a recent transformer-based crystal encoder, developing a new encoder architecture called CrystalFramer. Extensive comparisons with conventional frames and crystal encoders show the superior performance of the proposed method in various crystal property prediction tasks.
Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 5622
Loading