ADAPT-QAOA with a classically inspired initial state

TL;DR

This paper introduces a modified ADAPT-QAOA algorithm that uses a classical approximation-inspired initial state, leading to faster convergence and reduced likelihood of local minima in solving combinatorial optimization problems.

Contribution

The authors propose a novel initialization strategy for ADAPT-QAOA based on classical approximation solutions, improving convergence speed and solution quality.

Findings

Achieves same accuracy with fewer layers than standard QAOA and ADAPT-QAOA.

Less prone to converging to excited states and local minima.

Numerical simulations demonstrate improved performance.

Abstract

Quantum computing may provide advantage in solving classical optimization problems. One promising algorithm is the quantum approximate optimization algorithm (QAOA). There have been many proposals for improving this algorithm, such as using an initial state informed by classical approximation solutions. A variation of QAOA called ADAPT-QAOA constructs the ansatz dynamically and can speed up convergence. However, it faces the challenge of frequently converging to excited states which correspond to local minima in the energy landscape, limiting its performance. In this work, we propose to start ADAPT-QAOA with an initial state inspired by a classical approximation algorithm. Through numerical simulations we show that this new algorithm can reach the same accuracy with fewer layers than the standard QAOA and the original ADAPT-QAOA. It also appears to be less prone to the problem of converging to excited states.