IEEE International Symposium on Information Theory (ISIT)
Identifying Highly Distillable Entanglement in Pure States Using Multi-Armed Bandits
R Jayaharish, K. Bharati, Krishna Jagannathan
Abstract
The quantum internet uses high-quality entanglement distribution to enable long-distance communication. The quality of entanglement is measured by distillable entanglement, which quantifies how many maximally entangled states can be refined from a given ensemble of noisy states. Therefore, identifying highly distillable states is essential for ensuring efficient entanglement distribution. Given$K$ensembles of noisy entangled states, we address the problem of identifying highly distillable states, i.e., states with distillable entanglement exceeding a prespecified threshold. We propose entanglement distillation-based estimation protocols on two copies of bipartite$2 \times 2$and$2 \times 3$pure states. We characterize their distillable entanglement by estimating statedependent parameters from measurement statistics. We identify a structural connection to the Thresholding Bandit problem (TBP) in classical Multi-Armed Bandits (MAB) literature and use these policies to obtain guarantees on the sample complexity for estimating the state-dependent parameters. We also provide numerical results showcasing the performance of the proposed distillation protocol and TBP policies.