QUBO-Compatible Active Learning for Inverse Design of High-Entropy Alloys
Abstract
Machine-learned forward models can rapidly predict alloy properties, but their use for inverse design remains challenging when the search should also retain compatibility with quadratic unconstrained binary optimization (QUBO). Here, we develop a QUBO-compatible active-learning framework for inverse design of high-entropy alloys using a pretrained graph-neural-network predictor as a fixed property oracle. A property-guided binary variational autoencoder provides a binary latent representation, while an ensemble of quadratic factorization machines guides candidate selection. We systematically benchmark the framework through controlled latent-space ablations and comparison with direct composition-space optimization. The results show that candidate generation is a major determinant of search performance: local perturbations around previously high-performing latent codes provide the largest workflow-specific improvement, while surrogate-based selection further prioritizes candidates within the enriched search pool. The resulting QUBO-compatible workflow remains competitive with strong classical optimization strategies, although a composition-space genetic algorithm achieves the highest mean score. Finally, the learned quadratic surrogate can be exported directly as a QUBO. These results show that effective data acquisition can be separated from the final QUBO optimization endpoint, providing a benchmarked route for QUBO-compatible data-driven materials inverse design.
Related
- Inverse design of bespoke interatomic potentials via active learning by information-matching
- Quantum spectral thermodynamics and active learning enable million-scale exploration of high-entropy ceramics
- Symmetry- and Property-Aware Crystal Generation with Reinforcement Learning for Inverse Materials Design
- Deep learning enables large-scale inverse design of free-form metasurfaces
- Physics-guided inverse design of Co-based superalloys using machine learning and multi-objective optimization for enhanced γ' solvus temperature
BibTeX
@misc{silvi2026qubocompatible,
title = {{QUBO-Compatible Active Learning for Inverse Design of High-Entropy Alloys}},
author = {Giorgio Silvi and Kirsten Bark and Rolando Reiner and Nicolas Vogt and Thomas Plehn and Daniel Barragan-Yani and Marc Landmann and David Melching},
howpublished = {arXiv},
year = {2026},
month = aug,
doi = {10.48550/arxiv.2608.28239},
eprint = {2608.28239},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.28239},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - QUBO-Compatible Active Learning for Inverse Design of High-Entropy Alloys
AU - Giorgio Silvi
AU - Kirsten Bark
AU - Rolando Reiner
AU - Nicolas Vogt
AU - Thomas Plehn
AU - Daniel Barragan-Yani
AU - Marc Landmann
AU - David Melching
PY - 2026
DA - 2026/08/28
PB - arXiv
DO - 10.48550/arxiv.2608.28239
UR - https://arxiv.org/abs/2608.28239
AB - Machine-learned forward models can rapidly predict alloy properties, but their use for inverse design remains challenging when the search should also retain compatibility with quadratic unconstrained binary optimization (QUBO). Here, we develop a QUBO-compatible active-learning framework for inverse design of high-entropy alloys using a pretrained graph-neural-network predictor as a fixed property oracle. A property-guided binary variational autoencoder provides a binary latent representation, while an ensemble of quadratic factorization machines guides candidate selection. We systematically benchmark the framework through controlled latent-space ablations and comparison with direct composition-space optimization. The results show that candidate generation is a major determinant of search performance: local perturbations around previously high-performing latent codes provide the largest workflow-specific improvement, while surrogate-based selection further prioritizes candidates within the enriched search pool. The resulting QUBO-compatible workflow remains competitive with strong classical optimization strategies, although a composition-space genetic algorithm achieves the highest mean score. Finally, the learned quadratic surrogate can be exported directly as a QUBO. These results show that effective data acquisition can be separated from the final QUBO optimization endpoint, providing a benchmarked route for QUBO-compatible data-driven materials inverse design.
ER -