analiza aiqbench
The paper uses the trained neural tangent kernel to make a materials-property network interpretable, recovering a low-dimensional geometry that groups superconductors into recognizable families. Its main contribution is a practical method for probing learned representations in scientific ML, not a new superconductor. Readers should note that the evidence is a single model on elemental-composition data, so how well the kernel analysis generalizes to other architectures and materials tasks remains open.
Learning the Geometry of Materials Space with Trained Neural Tangent Kernels
arXiv:2610.27412v1 Announce Type: cross Abstract: Neural networks trained to predict materials properties learn representations whose physical content can be difficult to extract. We use the neural tangent kernel (NTK), evaluated at trained parameters, to recover the geometry learned by a network that predicts superconducting transition temperatures from elemental composition alone. Kernel ridge regression with the trained NTK closely reproduces the network's predictive performance and substantially outperforms regression with the initial kernel. Spectral analysis and multidimensional scaling reveal a geometry dominated by a small number of directions. Clustering and local neighborhood analysis organize compounds into recognizable superconducting families and connect materials with related chemical and correlated-electron properties despite substantial differences in $T_c$. These results establish trained tangent kernels as practical probes of learned materials representations, connecting property prediction with the exploration of materials space.
Abstract:Neural networks trained to predict materials properties learn representations whose physical content can be difficult to extract. We use the neural tangent kernel (NTK), evaluated at trained parameters, to recover the geometry learned by a network that predicts superconducting transition temperatures from elemental composition alone. Kernel ridge regression with the trained NTK closely reproduces the network's predictive performance and substantially outperforms regression with the initial kernel. Spectral analysis and multidimensional scaling reveal a geometry dominated by a small number of directions. Clustering and local neighborhood analysis organize compounds into recognizable superconducting families and connect materials with related chemical and correlated-electron properties despite substantial differences in $T_c$. These results establish trained tangent kernels as practical probes of learned materials representations, connecting property prediction with the exploration of materials space.
Materials Science (cond-mat.mtrl-sci); Superconductivity (cond-mat.supr-con); Machine Learning (cs.LG)
arXiv:2610.27412 [cond-mat.mtrl-sci]
(or arXiv:2610.27412v1 [cond-mat.mtrl-sci] for this version)
https://doi.org/10.48550/arXiv.2610.27412 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 227 cuvinte
Subiecte conexe
Open Materials Generation with Inference-Time Reinforcement Learning
arXiv · 12 iun. 2026
Robust and Efficient AI Frameworks for Scalable Material Design and Property Prediction
arXiv · 17 sept. 2026
Transfer Learning Architectures for Scalable Multi-Fidelity Bayesian Optimization
arXiv · 28 iul. 2026
Framework-Conditional Open Materials Generation
arXiv · 30 oct. 2026