Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.02818 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911136067092480 |
|---|---|
| author | Koroteev, Grigory |
| author_facet | Koroteev, Grigory |
| contents | In this paper, we introduce an extension of the concept of Van Hoff singularities in order to explicitly use them in machine learning methods of models as a priori information. The claimed method becomes possible when, instead of directly calculating the density of quantum states, we operate with estimates obtained by machine learning methods. Then, determining additional information allows us to organize post-processing, which significantly improves the quality of model prediction. Analytically, the effect is similar to Van Hove singularities of high order due to the additional degree of degeneracy of the system. As a mechanism for redistributing information about the system, we propose regularization, which is close in meaning to taking into account Fisher information. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_02818 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Quasi-van Hove singularities informed approach improving DOS/pDOS predictions in GNN Koroteev, Grigory Quantum Physics In this paper, we introduce an extension of the concept of Van Hoff singularities in order to explicitly use them in machine learning methods of models as a priori information. The claimed method becomes possible when, instead of directly calculating the density of quantum states, we operate with estimates obtained by machine learning methods. Then, determining additional information allows us to organize post-processing, which significantly improves the quality of model prediction. Analytically, the effect is similar to Van Hove singularities of high order due to the additional degree of degeneracy of the system. As a mechanism for redistributing information about the system, we propose regularization, which is close in meaning to taking into account Fisher information. |
| title | Quasi-van Hove singularities informed approach improving DOS/pDOS predictions in GNN |
| topic | Quantum Physics |
| url | https://arxiv.org/abs/2509.02818 |