Controllable Memory in Quantum Reservoir Networks

Researchers advance recurrent quantum reservoir computing architectures by introducing tunable partial-SWAP operations for controllable memory capacity. The study identifies feedback-based and recurrent models as two major competing architectures in QRC, with recurrent models using a two-register approach for fading memory. While these recurrent models have been validated on NISQ QPUs and shown real-world performance, the exact mechanism of memory capacity was not fully understood or controllable. The new work augments recurrent approaches, enabling control over memory through partial-SWAP gates, and presents a hardware-implementable design.