Paper ReviewMathematics & StatisticsOptimization & Operations Research

Max-Cut: Where Quantum Optimization Meets Classical Approximation on Real Graphs

The Max-Cut problem—partitioning a graph to maximize edges between partitions—is NP-hard and a proving ground for quantum advantage. Tate & Gupta identify small graph families where QAOA outperforms the classical Goemans-Williamson algorithm, providing concrete instances for near-term quantum benchmarking.

By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.

The Max-Cut problem asks a simple question: given a graph, partition its vertices into two groups to maximize the number of edges between groups. The problem is NP-hard—no known classical algorithm solves it optimally in polynomial time for general graphs. The best classical approximation, the Goemans-Williamson (GW) algorithm using semidefinite programming and hyperplane rounding, achieves an approximation ratio of approximately 0.878—meaning it finds a cut that is at least 87.8% as large as the optimal cut.

The Quantum Approximate Optimization Algorithm (QAOA), proposed by Farhi et al. in 2014, offers a quantum approach to Max-Cut and related combinatorial optimization problems. QAOA uses parametrized quantum circuits to prepare quantum states that encode approximate solutions, then classically optimizes the circuit parameters to maximize the expected cut value.

The central question—does QAOA outperform classical approximation algorithms?—has been surprisingly difficult to answer. Tate & Gupta address this by identifying specific graph families where QAOA shows advantage over GW hyperplane rounding, providing the concrete instances needed for near-term quantum benchmarking.

The Search for Quantum Advantage Instances

Demonstrating quantum advantage requires finding problems where quantum computers outperform the best classical algorithms. For QAOA on Max-Cut, this means finding graphs where QAOA achieves a better approximation ratio than GW's 0.878 guarantee.

Tate & Gupta's approach is systematic: they enumerate regular graph families (graphs where every vertex has the same degree) up to a few hundred vertices, compute both the QAOA expected cut and the GW approximation for each, and identify families where QAOA wins.

Their key findings:

  • On most graph families, GW hyperplane rounding and QAOA perform comparably
  • On specific regular graph families with particular structural properties, QAOA achieves measurably better approximation ratios
  • These advantageous instances tend to be small (under 1000 vertices)—important because they are within reach of near-term NISQ (noisy intermediate-scale quantum) devices
The identification of small, challenging instances is practically valuable: it provides concrete benchmarks for quantum hardware development. Rather than testing quantum computers on random graphs (where classical algorithms usually suffice), researchers can focus on these structurally specific instances where quantum advantage is theoretically predicted.

Graph Cuts in Practice

Singh & Saha provide a broader survey of graph cut problems in QAOA, covering not only Max-Cut but also related problems: minimum cut, balanced partitioning, and graph bisection. Their contribution is connecting the abstract QAOA framework to practical applications:

  • VLSI design: Partitioning circuit components to minimize wire crossings
  • Network analysis: Identifying community structure in social networks
  • Data clustering: Partitioning data into coherent groups
  • Image segmentation: Partitioning pixels into foreground and background
For each application, the relevant graph cut variant has different structural properties—and the relative advantage of QAOA vs. classical methods may differ. Understanding which problem structures favor quantum approaches is essential for identifying the most productive early applications of quantum optimization.

Claims and Evidence

<
ClaimEvidenceVerdict
QAOA outperforms GW on specific graph familiesTate & Gupta identify concrete advantageous instances✅ Supported (specific instances)
QAOA provides general advantage over classical Max-Cut algorithmsAdvantage is instance-dependent, not universal❌ Not universal
Small advantageous instances are within NISQ device reachIdentified instances have under 1000 vertices✅ Supported
Graph cut optimization has broad practical applicationsApplications in VLSI, networking, clustering well-documented✅ Well-established
Current quantum hardware can realize QAOA advantageNoise and error rates limit practical implementation⚠️ Approaching but not yet demonstrated

Open Questions

  • Noise resilience: QAOA on real quantum hardware is affected by gate errors and decoherence. Do the theoretically advantageous instances retain their advantage under realistic noise models?
  • Scaling: The advantage instances identified are small. Does the quantum advantage persist, grow, or diminish as graph size increases? Theoretical analysis suggests the answer depends on the QAOA depth (number of circuit layers).
  • Parameter optimization: QAOA requires classically optimizing circuit parameters—a non-convex optimization problem that may have many local minima. Do parameter optimization challenges limit practical QAOA performance?
  • Beyond Max-Cut: QAOA applies to any combinatorial optimization problem expressible as a quadratic unconstrained binary optimization (QUBO). For which QUBO problems is quantum advantage most likely?
  • Classical competition: Classical algorithms continue to improve. Will classical advances (better approximation algorithms, smarter heuristics) close the gap before quantum hardware becomes practical?
  • What This Means for Your Research

    For quantum computing researchers, the identification of specific advantageous instances provides concrete targets for hardware benchmarking. Demonstrating QAOA advantage on these instances—even small ones—would be a meaningful milestone toward practical quantum optimization.

    For optimization researchers, the QAOA-classical comparison illuminates the structural properties of problems that favor quantum approaches. Understanding why certain graph families are harder for classical algorithms may lead to improved classical methods as well.

    For practitioners in VLSI design, network analysis, and data science who use graph partitioning, the quantum optimization horizon is relevant for long-term planning—but classical methods remain the practical choice for current applications.

    References (2)

    [1] Tate, R. & Gupta, S. (2025). Comparison of Hyperplane Rounding for Max-Cut and Quantum Approximate Optimization Algorithm over Certain Regular Graph Families. arXiv:2509.24108.
    [2] Singh, N. & Saha, A. (2025). Exploring graph cuts in quantum approximate optimization algorithm. Physica Scripta.

    Explore this topic deeper

    Search 290M+ papers, detect research gaps, and find what hasn't been studied yet.

    Click to remove unwanted keywords

    Search 8 keywords →