Core Concepts
Error mitigation strategies may not fully resolve exponential cost concentration issues in noisy VQAs, leading to potential trainability challenges.
Abstract
The study explores the impact of error mitigation on noisy Variational Quantum Algorithms (VQAs). It investigates various strategies like Zero-Noise Extrapolation, Virtual Distillation, Probabilistic Error Cancellation, and Clifford Data Regression. Results suggest that while some methods can aid training, others may worsen trainability due to noise-induced barriers.
VQAs are crucial for quantum advantage.
Noise limits trainability by flattening cost landscapes.
Error Mitigation aims to reduce noise impact.
Different EM strategies have varying effects on resolving cost function values.
Careful selection of EM methods is essential for improving VQA trainability.
Stats
Noise can severely limit the trainability of VQAs.
Exponential resources are needed to resolve cost concentration issues with certain EM strategies.
Quotes
"EM protocols should be carefully applied as they can either worsen or not improve trainability." - Authors