Near-term optimization and machine learning applications in emerging superconducting qudit processors

We discuss progress toward the design of near-term applications of superconducting qudit processors based on superconducting cavity–transmon modules under development by multiple labs and companies. These multi-level architectures provide a natural platform for exploring hybrid optimization and machine-learning schemes that exploit large Hilbert spaces with minimal low depth [1]. We outline the functional use of amplitude damping noise via the Noise-Directed Adaptive Remapping (NDAR) technique [2] for combinatorial optimization with qudits, as well as reservoir-computing real-time signal processing protocols using coupled cavity modes as analog quantum reservoirs. Preliminary work suggests that measurable scaling advantages from higher-dimensional encodings and open-system dynamics could be discovered. Ongoing experiments on a two-qubit system developed at SQMS are testing these design directly.