Neural-Network Inverse Design of SRF Cavities and Transmons for Bosonic Quantum Computation

The paper presents two deep neural-network approaches for inverse design in bosonic quantum hardware. One predicts three-dimensional superconducting radio-frequency cavity geometries from target cavity observables. The other predicts transmon designs from target qubit-cavity coupling parameters: coupling rate, qubit frequency, and anharmonicity (g, ν_q, α). End-to-end re-simulation reportedly shows approximately 5% target error for cavity designs and approximately 2% for transmon designs. The method addresses the one-to-many nature of inverse design and aims to reduce the cost of conventional iterative electromagnetic simulation as device design spaces expand.