Noise-informed parameter prediction

Determine whether a noise-informed parameter-prediction method based on the error-resilience bound improves QAOA performance beyond parameter-generation methods that use unoptimized predicted parameters without incorporating noise.

Background

The paper finds that parameters generated without optimization can outperform parameters optimized without noise when QAOA is executed under noise. This motivates using the derived parameter-dependent error-resilience bound to design parameter-prediction methods that account explicitly for noise.

The authors leave unresolved whether such a noise-informed prediction strategy would provide further improvement over the parameter-generation approaches studied in the paper, particularly beyond the small numerical instances considered.