Reproduce “Adaptive Job Scheduling in Quantum Clouds Using Reinforcement Learning” on OpenResearch

About this paper

Waylon Luo, Jiapeng Zhao, Tong Zhan, Qiang Guan

Researchers from Kent State University, Cisco, and Meta developed the first open-source simulation framework for distributed quantum computing, which models partitioning large quantum circuits across multiple QPUs. This framework was used to evaluate four scheduling strategies, demonstrating that an error-aware approach achieves the highest average circuit fidelity (0.6878) at the cost of increased execution time, while a reinforcement learning strategy offered the fastest execution but with lower fidelity.

Key insights

• Distributed quantum computing is a feasible and necessary approach for scaling quantum algorithms beyond the limitations of single quantum processing units (QPUs).
• A fundamental trade-off exists between execution speed and quantum circuit fidelity in distributed systems, highlighting the need for sophisticated resource allocation strategies.
• Inter-device classical communication introduces measurable fidelity degradation, emphasizing the importance of minimizing partitioning and having low-latency communication channels.
• Reinforcement learning-based scheduling shows promise for optimizing execution time but requires further refinement to effectively balance fidelity objectives in complex quantum cloud environments.

Summary by alphaXiv.