Zero-shot Quantum Neural Architecture Search

2026-05-12

AI Summary

This work addresses the challenge of designing high-performing variational quantum algorithms (VQAs), whose efficacy hinges on quantum circuit architecture and requires careful trade-offs among expressivity, trainability, and hardware constraints. Existing neural architecture search methods incur prohibitive computational costs due to repeated training of candidate circuits. To overcome this, the study introduces zero-shot performance estimation into quantum neural architecture search for the first time, leveraging the convergence properties of Gram matrices derived from the quantum neural tangent kernel to construct a surrogate model that evaluates architectures without full training. Coupled with Monte Carlo tree search for efficient exploration of the architectural space, the proposed MZeQAS framework significantly outperforms existing approaches across multiple tasks, simultaneously enhancing both search efficiency and final model performance, thereby offering a scalable solution for deploying VQAs on noisy intermediate-scale quantum devices.

Abstract

Variational Quantum Algorithms (VQAs) are a leading approach to exploiting near-term quantum hardware, leveraging parameterized quantum circuits and classical optimization to achieve advantage. Despite their promise, the practical deployment of VQAs is challenged by the difficulty of designing quantum circuit architectures that balance expressivity, trainability, and hardware constraints. Existing evolutionary-based quantum neural architecture search methods address these challenges but suffer from high computational costs due to repeated training of candidate circuits. In this work, we identify a setting in which the Gram matrix of the Quantum Neural Tangent Kernel converges. Building on this observation, we design a zero-shot surrogate model to estimate candidate performance without full training, significantly accelerating the architecture search process. Using this surrogate, we propose MZeQAS, a Monte Carlo Tree Search (MCTS)-based Zero-Shot Quantum Neural Architecture Search framework for VQAs. By integrating proxy-based performance estimation with MCTS exploration, MZeQAS efficiently discovers high-performing architectures. Experimental results demonstrate that MZeQAS outperforms existing approaches in terms of both search efficiency and solution quality, providing a scalable and effective framework for advancing VQA deployment on noisy intermediate-scale quantum devices.

Problem

• Quantum Neural Architecture Search
• Variational Quantum Algorithms
• Zero-shot
• Quantum Circuit Design
• Computational Efficiency

Innovation

• Zero-shot
• Quantum Neural Architecture Search
• Quantum Neural Tangent Kernel
• Monte Carlo Tree Search
• Variational Quantum Algorithms