Quantum optimisation via maximally amplified states

TL;DR

This paper introduces the MAOA, a quantum algorithm that amplifies optimal solutions for combinatorial problems more effectively than existing methods, enabling faster near-term quantum optimization.

Contribution

The paper presents the MAOA, a new quantum algorithm that maximizes solution amplification without deep circuits, outperforming QAOA and modified Grover search in simulations.

Findings

MAOA significantly amplifies optimal solutions compared to other algorithms.

MAOA achieves substantial speedup over classical sampling methods.

Numerical results demonstrate convergence to a theoretical upper bound.

Abstract

This paper presents the Maximum Amplification Optimisation Algorithm (MAOA), a novel quantum algorithm designed for combinatorial optimisation in the restricted circuit depth context of near-term quantum computing. The MAOA first produces a quantum state in which the optimal solutions to a problem are amplified to the maximum extent possible subject to a given restricted circuit depth. Subsequent repeated preparation and measurement of this maximally amplified state produces solutions of the highest quality as efficiently as possible. The MAOA performs considerably better than other near-term quantum algorithms, such as the Quantum Approximate Optimisation Algorithm (QAOA), as it amplifies optimal solutions significantly more and does so without the computationally demanding variational procedure required by these other algorithms. Additionally, a restricted circuit depth modification of the existing Grover adaptive search is introduced. This modified algorithm is referred to as the restricted Grover adaptive search (RGAS), and provides a useful comparison to the MAOA. The MAOA and RGAS are simulated on a practical vehicle routing problem, a computationally demanding portfolio optimisation problem, and an arbitrarily large problem with normally distributed solution qualities. In all cases, the MAOA and RGAS are shown to provide substantial speedup over classical random sampling in finding optimal solutions, while the MAOA consistently outperforms the RGAS. The speedup provided by the MAOA is quantified by demonstrating numerical convergence to a theoretically derived upper bound.