Noisy intermediate-scale quantum algorithms

Analysis

Why This Paper Matters

This review, published in Reviews of Modern Physics and with over 1,600 citations, is a landmark survey of the noisy intermediate-scale quantum (NISQ) era. It systematically maps out what is achievable with current quantum hardware, which is too noisy for full fault-tolerant quantum computing but capable of outperforming classical computers on certain tasks. The paper is essential reading for AI practitioners because it clarifies the realistic capabilities and limitations of quantum machine learning and optimization, helping to set expectations for near-term quantum advantage.

The paper matters because it consolidates a rapidly growing field into a coherent framework, identifying the key algorithmic paradigms—variational quantum eigensolvers (VQE), quantum approximate optimization algorithm (QAOA), and quantum machine learning models—that dominate the NISQ landscape. It also highlights the importance of error mitigation and hybrid classical-quantum approaches, which are crucial for making NISQ devices useful.

Technical Contributions

• Comprehensive taxonomy of NISQ algorithms: The paper categorizes algorithms by application (physics, chemistry, optimization, machine learning) and by computational approach (variational, adiabatic, sampling-based).
• Detailed discussion of VQE: Explains how variational quantum circuits can approximate ground states of molecular Hamiltonians, including hardware-efficient ansätze and problem-specific encodings.
• Analysis of QAOA: Covers the algorithm's structure, performance guarantees, and extensions for constrained optimization problems.
• Quantum machine learning models: Reviews quantum neural networks, kernel methods, and generative models, noting their potential for data classification and feature mapping.
• Error mitigation techniques: Summarizes methods like zero-noise extrapolation, probabilistic error cancellation, and measurement error mitigation that improve NISQ device outputs.

Results

The review does not present new experimental results but synthesizes findings from numerous studies. Key takeaways include: VQE has been demonstrated on small molecules (e.g., H2, LiH) with up to 12 qubits, achieving chemical accuracy with error mitigation. QAOA has shown performance comparable to classical heuristics on MaxCut problems with up to 20 qubits. Quantum machine learning models have achieved classification accuracy on small datasets (e.g., handwritten digits) but have not yet demonstrated a clear quantum advantage. The paper emphasizes that current NISQ devices have error rates around 10^-3 per gate, limiting circuit depth to about 100 gates before noise dominates.

Significance

This review has shaped the research agenda for quantum computing in the NISQ era, influencing both theoretical work and experimental implementations. For AI, it provides a sobering assessment: while quantum machine learning is an active area, practical quantum advantage for AI tasks remains elusive without fault-tolerant hardware. The paper's emphasis on hybrid classical-quantum algorithms has spurred development of software frameworks (e.g., PennyLane, Qiskit) that integrate quantum circuits into classical machine learning pipelines. Its high citation count reflects its role as a definitive reference for researchers entering the field.