DAPO-QAOA: An algorithm for solving combinatorial optimization problems by dynamically constructing phase operators

TL;DR

DAPO-QAOA introduces a dynamic method for constructing phase operators in QAOA, improving approximation ratios and reducing two-qubit gate usage, making it more suitable for NISQ devices.

Contribution

The paper proposes DAPO-QAOA, a novel adaptive algorithm that constructs phase operators based on previous outputs, enhancing efficiency and performance in combinatorial optimization.

Findings

Achieves higher approximation ratios on MaxCut and NAE3SAT.

Reduces two-qubit RZZ gates by 34% compared to vanilla QAOA.

Demonstrates effectiveness especially in dense graphs.

Abstract

The Quantum Approximate Optimization Algorithm (QAOA) is a well-known hybrid quantum-classical algorithm for combinatorial optimization problems. Improving QAOA involves enhancing its approximation ratio while addressing practical constraints of Noisy Intermediate Scale Quantum (NISQ) devices, such as minimizing the number of two-qubit gates and reducing circuit depth. Although existing research has optimized designs for phase and mixer operators to improve performance, challenges remain, particularly concerning the excessive use of two-qubit gates in the construction of phase operators. To address these issues, we introduce a Dynamic Adaptive Phase Operator (DAPO) algorithm, which dynamically constructs phase operators based on the output of previous layers and neighborhood search approach, optimizing the problem Hamiltonian more efficiently. By using solutions generated by QAOA itself to simplify the problem Hamiltonian at each layer, the algorithm captures the problem's structural properties more effectively, progressively steering the solution closer to the optimal target. Experimental results on MaxCut and NAE3SAT problems show that DAPO achieves higher approximation ratios and significantly reduces two-qubit RZZ gates, especially in dense graphs. Compared to vanilla QAOA, DAPO uses only 66% of RZZ gates at the same depth while delivering better results, demonstrating its potential for efficient combinatorial optimization in the NISQ era.