Dynamic local operations and classical communication for automated entanglement manipulation

Distributed quantum computing offers a promising pathway to overcome the limitations of individual quantum processors by connecting them into a networked system. Due to the physical constraints, the allowed quantum operation in the distributed quantum computing paradigm is local operations and classical communication (LOCC). However, designing practical LOCC protocols for large systems is generally challenging, often requiring exponential computational resources. Here, we propose a general and flexible framework called dynamic LOCCNet (DLOCCNet) for designing and optimizing LOCC protocols using optimization techniques. Rather than designing large-scale protocols directly, DLOCCNet decomposes the large-size problems into small, recursively trainable optimization problems. Protocols designed by this framework achieve performance comparable to existing methods while significantly reducing computational resource demands. We conduct numerical experiments to demonstrate its effectiveness in entanglement distillation and distributed state discrimination tasks. The authors propose a general and flexible framework called Dynamic LOCCNet for designing and optimizing LOCC protocols. Their method decomposes large-scale problems into smaller, recursively trainable optimization problems, achieving comparable performance to existing methods while significantly reducing computational resource demands, making it a practical and scalable tool for current quantum devices.