AutoML-driven optimization of variational quantum circuit

Haozhen Situ, Zhengjiang Li, Zhimin He, Qin Li, Jinjing Shi

Abstract: Highlights•We optimize both circuit structures and training settings for quantum architecture search.•A graph convolutional network predicts performance from circuit structures and settings.•Novel encoding handles varied gate counts and feature heterogeneity in quantum circuits.•Self-supervised learning with masked gate reconstruction reduces training complexity.•Simulations show our approach efficiently finds high-performance quantum circuits.