[Paper Review] Quantum Approximate Optimization Algorithm: Performance, Mechanism, and Implementation on Near-Term Devices

Leo Zhou, Sheng-Tao Wang+3 more|arXiv (Cornell University)|Dec 3, 2018

Quantum Computing Algorithms and Architecture55 citations

TL;DR

The paper develops efficient heuristic parameter optimization for QAOA, benchmarks its performance on MaxCut, compares it with quantum annealing, and proposes a scalable route for near-term experimental implementation with neutral atoms.

ABSTRACT

The Quantum Approximate Optimization Algorithm (QAOA) is a hybrid quantum-classical variational algorithm designed to tackle combinatorial optimization problems. Despite its promise for near-term quantum applications, not much is currently understood about QAOA's performance beyond its lowest-depth variant. An essential but missing ingredient for understanding and deploying QAOA is a constructive approach to carry out the outer-loop classical optimization. We provide an in-depth study of the performance of QAOA on MaxCut problems by developing an efficient parameter-optimization procedure and revealing its ability to exploit non-adiabatic operations. Building on observed patterns in optimal parameters, we propose heuristic strategies for initializing optimizations to find quasi-optimal $p$-level QAOA parameters in $O( ext{poly}(p))$ time, whereas the standard strategy of random initialization requires $2^{O(p)}$ optimization runs to achieve similar performance. We then benchmark QAOA and compare it with quantum annealing, especially on difficult instances where adiabatic quantum annealing fails due to small spectral gaps. The comparison reveals that QAOA can learn via optimization to utilize non-adiabatic mechanisms to circumvent the challenges associated with vanishing spectral gaps. Finally, we provide a realistic resource analysis on the experimental implementation of QAOA. When quantum fluctuations in measurements are accounted for, we illustrate that optimization will be important only for problem sizes beyond numerical simulations, but accessible on near-term devices. We propose a feasible implementation of large MaxCut problems with a few hundred vertices in a system of 2D neutral atoms, reaching the regime to challenge the best classical algorithms.

Motivation & Objective

• Investigate how to efficiently optimize QAOA parameters beyond p=1 for MaxCut problems.
• Characterize the performance of QAOA at intermediate depths and identify mechanisms beyond adiabatic evolution.
• Develop and test heuristic initialization strategies to achieve near-optimal QAOA performance in polynomial time.
• Compare QAOA with quantum annealing to understand non-adiabatic learning and potential advantages.
• Provide a realistic resource assessment and outline feasible near-term experimental implementations.

Proposed method

• Formulate QAOA as a p-level variational quantum circuit with 2p parameters and maximize the objective via classical optimization.
• Identify and exploit patterns in optimal parameters to develop heuristics for initial points (INTERP and FOURIER).
• Introduce a Discrete Sine/Cosine Transform parametrization to reduce optimization dimensionality.
• Benchmark QAOA on random u3R and w3R MaxCut graphs up to p≤50 and N≤22, using gradient-based optimizers like BFGS.
• Compare QAOA performance to adiabatic quantum annealing, focusing on hard instances with small spectral gaps.
• Discuss experimental resource considerations, including measurement noise and a proposed 2D neutral-atom implementation.

Experimental results

Research questions

• RQ1 Can QAOA at intermediate depth (p>1) outperform classical algorithms on MaxCut graphs?
• RQ2 How can one efficiently optimize QAOA parameters beyond random initialization?
• RQ3 Does QAOA leverage non-adiabatic mechanisms to overcome small spectral gaps in comparison to quantum annealing?
• RQ4 What are the practical resource requirements and feasibility for implementing large MaxCut problems on near-term devices?

Key findings

• Optimal QAOA parameters exhibit persistent, smoothly varying patterns with increasing p, enabling efficient initialization.
• Heuristic strategies (INTERP and FOURIER) yield quasi-optimal parameters in polynomial time, outperforming random initialization which scales exponentially.
• QAOA can outperform adiabatic quantum annealing on hard instances with small spectral gaps by utilizing non-adiabatic mechanisms.
• For u3R graphs, average fractional error decays roughly as a stretched exponential with p; for w3R graphs, decay is approximately exponential in sqrt(p).
• Measurement fluctuations (projection noise) suggest optimization becomes more critical only for very large problem sizes, but near-term devices can still demonstrate practical advantage with hundreds of vertices in a 2D neutral-atom setup.