Noise-induced shallow circuits and the absence of barren plateaus

Key findings from this study

This research indicates that:

• Local noise truncates quantum circuits to logarithmic depth, preventing deep quantum computation for observable expectation value estimation.
• Non-unital noise eliminates barren plateaus for cost functions based on local observables.
• Classical algorithms can efficiently simulate noisy quantum circuits within constant additive accuracy across arbitrary architectures.

Overview

This theoretical study investigates how uncorrected local noise affects quantum circuit performance, focusing on observable expectation value estimation. The work establishes that non-unital noise prevents barren plateaus while simultaneously rendering most circuits effectively shallow, enabling efficient classical simulation.

Methods and approach

The authors prove theoretically that local noise truncates quantum circuits to logarithmic depth and demonstrate that non-unital noise precludes barren plateau phenomena. They subsequently design a classical algorithm that exploits this effective shallowness to estimate observable expectation values within constant additive accuracy across diverse circuit architectures.

Results

Local noise mechanisms effectively compress quantum circuits to logarithmic depth when estimating observable expectation values. This compression occurs across any circuit architecture subject to non-unital noise. The authors proved that cost functions composed of local observables never exhibit barren plateaus under such noise conditions. Simultaneously, they constructed a classical algorithm capable of achieving constant additive accuracy in estimating observable expectation values with high probability, leveraging the circuits' inherent shallowness. These findings hold regardless of the specific circuit architecture employed.

Implications

The results indicate that noisy near-term quantum circuits lack inherent computational advantages for observable expectation value estimation unless specifically engineered to exploit noise properties. Variational quantum machine learning proposals relying on such estimation tasks face fundamental limitations when operating without error correction. This work suggests that meaningful quantum advantage in this domain requires either robust error correction capabilities or algorithmic designs that deliberately harness noise rather than operate despite it.

Scope and limitations

This summary is based on the study abstract and available metadata. It does not include a full analysis of the complete paper, supplementary materials, or underlying datasets unless explicitly stated. Findings should be interpreted in the context of the original publication.