Transformers as Intrinsic Optimizers for Quantum Approximate Optimization Algorithm
Abstract
The Quantum Approximate Optimization Algorithm (QAOA) is a leading variational framework for combinatorial optimization on noisy intermediate-scale quantum hardware, but its practical performance depends strongly on the classical optimizer used to train its variational parameters. This outer-loop optimization is often nonconvex, initialization-sensitive, and costly when repeated across large families of related problem instances. In this work, we propose a Transformer-based intrinsic optimization framework for QAOA, in which the optimizer itself is learned and embedded directly into the hybrid quantum-classical loop. The proposed graph-conditioned Transformer processes problem structure, current QAOA parameters, measurement feedback, and recent optimization history to predict the next variational-parameter update, thereby reformulating instance-wise classical optimization as an amortized learned policy. We develop a mathematical formulation of this intrinsic-optimization perspective and evaluate the method on QAOA-based MaxCut benchmarks across multiple problem settings, with comparisons against representative classical and learned optimization baselines. The results demonstrate that Transformer-based intrinsic optimization can provide a structured and transferable mechanism for improving the classical component of hybrid quantum optimization algorithms.
Related
- Depth analysis of the Quantum Approximate Optimization Algorithm with a Grover mixer
- κ-Parametrized Boltzmann Distribution Sampling on Quantum Approximate Optimization Algorithm
- Parameter-Dependent Noise Resilience in the Quantum Approximate Optimization Algorithm
- Symmetry-Preserving Hierarchical Quantum Approximate Optimization Algorithm with Subproblems Beyond the Qubit Limit
- Quantum Approximate Optimisation Algorithm for Protein Sidechain Packing
BibTeX
@misc{chen2026transformers,
title = {{Transformers as Intrinsic Optimizers for Quantum Approximate Optimization Algorithm}},
author = {Kuan-Cheng Chen and Xiaotian Xu and Hiromichi Matsuyama and Wei-Hao Huang and Haomu Yuan and Yu Yamashiro},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.19392},
eprint = {2609.19392},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.19392},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Transformers as Intrinsic Optimizers for Quantum Approximate Optimization Algorithm
AU - Kuan-Cheng Chen
AU - Xiaotian Xu
AU - Hiromichi Matsuyama
AU - Wei-Hao Huang
AU - Haomu Yuan
AU - Yu Yamashiro
PY - 2026
DA - 2026/09/16
PB - arXiv
DO - 10.48550/arxiv.2609.19392
UR - https://arxiv.org/abs/2609.19392
AB - The Quantum Approximate Optimization Algorithm (QAOA) is a leading variational framework for combinatorial optimization on noisy intermediate-scale quantum hardware, but its practical performance depends strongly on the classical optimizer used to train its variational parameters. This outer-loop optimization is often nonconvex, initialization-sensitive, and costly when repeated across large families of related problem instances. In this work, we propose a Transformer-based intrinsic optimization framework for QAOA, in which the optimizer itself is learned and embedded directly into the hybrid quantum-classical loop. The proposed graph-conditioned Transformer processes problem structure, current QAOA parameters, measurement feedback, and recent optimization history to predict the next variational-parameter update, thereby reformulating instance-wise classical optimization as an amortized learned policy. We develop a mathematical formulation of this intrinsic-optimization perspective and evaluate the method on QAOA-based MaxCut benchmarks across multiple problem settings, with comparisons against representative classical and learned optimization baselines. The results demonstrate that Transformer-based intrinsic optimization can provide a structured and transferable mechanism for improving the classical component of hybrid quantum optimization algorithms.
ER -