Stone-in-Waiting: A Cloud-Based Accelerator for the Quantum Approximate Optimization Algorithm

2026-03-20

AI Summary

This work addresses the challenge of parameter initialization in the Quantum Approximate Optimization Algorithm (QAOA) during the noisy intermediate-scale quantum (NISQ) era. To tackle this issue, the authors propose a novel initialization method that integrates Bayesian optimization, a nearest-neighbor strategy, and metric learning—the first such combination applied to QAOA. The approach is implemented within a cloud-native acceleration platform featuring a web interface and API support for usability and scalability. Experimental results demonstrate that the proposed method yields initial parameters that improve optimization performance by 40.19% compared to baseline approaches. Furthermore, comprehensive comparisons across multiple algorithms confirm the effectiveness and practical utility of the proposed initialization scheme in real-world quantum optimization scenarios.

Technology Category

• Machine Learning: Quantum Machine Learning
• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Constraint Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

The Quantum Approximate Optimization Algorithm (QAOA) and its advanced variant, the Quantum Alternating Operator Ansatz (QAOA), are major research topics in the current era of Noisy Intermediate-Scale Quantum (NISQ) computing. However, the problem of initializing their parameters remains unresolved. Motivated by the combinatorial optimization task in the 6th MindSpore Quantum Computing Hackathon (2024), this paper proposes Stone-in-Waiting, a cloud-based accelerator for obtaining high-quality initial parameters for QAOA. Internally, the accelerator builds on state-of-the-art theories and methods for parameter determination and integrates four self-developed algorithms for QAOA parameter initialization, mainly based on Bayesian methods, nearest-neighbor methods, and metric learning. Compared with the Baseline Algorithm, the generated parameters improve the score by 40.19%. Externally, the accelerator offers both a web interface and an API, providing flexible and convenient access for users to test and develop related experiments and applications. This paper presents the design principles and methods of Stone-in-Waiting, demonstrates its functional characteristics, compares the strengths and weaknesses of the four proposed algorithms, and validates the overall system performance through experiments.

Problem

• Quantum Approximate Optimization Algorithm
• parameter initialization
• NISQ
• combinatorial optimization
• quantum computing

Innovation

• QAOA
• parameter initialization
• Bayesian optimization
• cloud-based accelerator
• metric learning