![[2602.16908] Multi-objective optimization and quantum hybridization of equivariant deep learning interatomic potentials on organic and inorganic compounds](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.16908] Multi-objective optimization and quantum hybridization of equivariant deep learning interatomic potentials on organic and inorganic compounds
Summary
This article presents a study on the multi-objective optimization of deep learning interatomic potentials, focusing on the trade-off between accuracy and inference time in machine learning models for atomic property predictions.
Why It Matters
The research addresses critical challenges in materials science and quantum physics by improving the efficiency and accuracy of machine learning interatomic potentials, which are essential for simulating molecular interactions. This has implications for both organic and inorganic compound analysis, potentially accelerating advancements in material design and discovery.
Key Takeaways
- Introduces Allegro, a machine learning model for predicting atomic properties.
- Explores multi-objective optimization to balance accuracy and inference time.
- Tests various model architectures, including quantum-classical hybrids.
- Demonstrates improved accuracy over existing models with specific datasets.
- Highlights the significance of optimizing hyperparameters in ML models.
Condensed Matter > Materials Science arXiv:2602.16908 (cond-mat) [Submitted on 18 Feb 2026] Title:Multi-objective optimization and quantum hybridization of equivariant deep learning interatomic potentials on organic and inorganic compounds Authors:G. Laskaris, D. Morozov, D. Tarpanov, A. Seth, J. Procelewska, G. Sai Gautam, A. Sagingalieva, R. Brasher, A. Melnikov View a PDF of the paper titled Multi-objective optimization and quantum hybridization of equivariant deep learning interatomic potentials on organic and inorganic compounds, by G. Laskaris and 8 other authors View PDF HTML (experimental) Abstract:Allegro is a machine learning interatomic potential (MLIP) model designed to predict atomic properties in molecules using E(3) equivariant neural networks. When training this model, there tends to be a trade-off between accuracy and inference time. For this reason we apply multi-objective hyperparameter optimization to the two objectives. Additionally, we experiment with modified architectures by making variants of Allegro some by adding strictly classical multi-layer perceptron (MLP) layers and some by adding quantum-classical hybrid layers. We compare the results from QM9, rMD17-aspirin, rMD17-benzene and our own proprietary dataset consisting of copper and lithium atoms. As results, we have a list of variants that surpass the Allegro in accuracy and also results which demonstrate the trade-off with inference times. Comments: Subjects: Materials Science (cond-mat.mtrl-...
