Reproduce “Benchmarking simulated and physical quantum processing units using quantum and hybrid algorithms” on OpenResearch

About this paper

Mohammad Kordzanganeh, Markus Buchberger, Basil Kyriacou, Maxim Povolotskii, Wilhelm Fischer, Andrii Kurkin, Wilfrid Somogyi, Asel Sagingalieva and 2 more

Researchers from Terra Quantum AG and QMware AG systematically benchmarked simulated and physical quantum processing units using Quantum Neural Networks (QNNs) and Hybrid Quantum Neural Networks (HQNNs), evaluating runtime, accuracy, and cost. The study compared QMware's proprietary simulator stack against AWS Braket's simulators and several commercial QPUs (IonQ, OQC, Rigetti, IBM), demonstrating QMware's basiq simulator offered up to a 78% runtime reduction for circuits under 27 qubits and highlighting the high fidelity of IonQ's Harmony for small circuits while revealing prohibitive QPU training costs.

Key insights

• Performance advantages stem significantly from the synergistic interplay between optimized quantum software and underlying hardware infrastructure, rather than raw hardware power alone.
• Quantum simulators offer a more cost-effective and accurate solution for QNN development and training in the NISQ era, especially for small to medium-sized circuits, despite QPUs showing potential for exponential runtime advantages at larger scales.
• Hybrid classical-quantum architectures are vital for leveraging quantum advantage in the near term, effectively combining the strengths of both classical and quantum resources.

Summary by alphaXiv.