Interaction-Aware Embedding Optimization for Neutral-Atom Quantum Processors

Abstract

Neutral atom-based quantum computers have the ability to dynamically reshape the qubit register geometry between experimental runs. This capability enables different types of problems to be mapped directly onto the hardware, significantly enhancing the efficiency of solving, for instance, complex optimization and machine learning use-cases. However, this flexibility requires finding atom positions to implement target pairwise interactions, inducing a difficult nonconvex optimization for setting up the quantum machine. In this work, we introduce a new force-directed method that targets interactions to encode the input problem on the quantum computer. Numerical results show that our new method can improve embedding feasibility and solution quality by orders of magnitude, and remains reliable on the largest instances.

Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.