Speaker - Edwardson School of Industrial Engineering - Purdue University

Artificial intelligence meets quantum state reconstruction

We present our quantum state tomography framework where state reconstruction is performed using artificial intelligence (AI) directly from a set of measurements. Our hardware-aware data-centric AI techniques reconstruct quantum states of comparable fidelity to that of a typical reconstruction method with the advantage that costly computations are front-loaded with our reconstructing setup. AI has found broad applicability in quantum information science, where existing AI techniques are often applied without significant alterations to network architectures. In this presentation, we demonstrate physics-inspired data-centric heuristics for AI systems used in quantum information science and their efficacy for quantum state reconstruction. Moreover, we discuss methods for enhancing the accuracy of our systems reconstruction by developing custom data sets that reflect essential properties, such as mean purity, of quantum systems we expect to encounter in experiments. Finally, we present custom prior distributions that are automatically tuned and generally better conform to the physical properties of the underlying system than standard fixed prior distributions in Bayesian quantum state estimation. Using both simulated and experimental measurement results, we show that AI-defined prior distributions reduce net convergence times and provide a natural way to incorporate implicit and explicit information directly into the prior distribution.