Generalization of risk-seeking optimization across state-preparation settings

Investigate whether the behaviors observed for risk-seeking optimization of noisy SNAP-Displacement state preparation generalize to different target states, larger Fock-space dimensions, deeper SNAP-Displacement ansatzes, and different noise models.

Background

The paper introduces trajectory-wise entropic-risk objectives for optimizing SNAP-Displacement ansatzes under photon-loss noise. In the reported experiments, risk-seeking optimization substantially lowers the minimum pure-state infidelity relative to average-fidelity optimization and provides a controllable trade-off among the infidelity threshold, the probability of attaining that threshold, and the optimization success rate.

These findings are established only for a specific experimental configuration: Haar-random target states, an eight-level Fock space, a four-block SNAP-Displacement ansatz, and the specified photon-loss noise model. The authors explicitly leave unresolved whether the observed improvements and trade-offs persist when the target states, Hilbert-space dimension, ansatz depth, or noise model are changed.