Rethinking Quantum Noise in Quantum Machine Learning: When Noise Improves Learning

TL;DR

This paper provides numerical evidence that quantum noise can sometimes improve the performance of quantum machine learning models, especially for poorly initialized models, challenging the traditional view of noise as purely detrimental.

Contribution

The study reveals that quantum noise can act as an implicit regularizer in quantum machine learning, with effects depending on initialization and model structure, suggesting new optimization strategies.

Findings

Approximately one-third of models improve under moderate noise

Negative correlation between baseline performance and noise benefit

Optimal noise level is below theoretical predictions

Abstract

Quantum noise is conventionally viewed as a fundamental obstacle in near-term quantum computing, motivating extensive error correction and mitigation strategies. We present numerical evidence that challenges this consensus. Through experiments on quantum graph neural networks for molecular property prediction, we discover that quantum noise induces heterogeneous, initialization-dependent responses. Among randomly initialized models with identical architecture, approximately one-third show performance improvement under moderate noise, while a smaller fraction deteriorate and the remainder are marginally affected. We identify a strong negative correlation ($r = -0.62$) between baseline model performance and noise benefit, suggesting that noise acts as an implicit regularizer for under-optimized models while disrupting well-converged ones. The observed optimal noise level falls below theoretical predictions, indicating error cancellation in structured quantum circuits. These findings demonstrate that quantum noise effects depend critically on initialization quality and need not be uniformly detrimental, suggesting a shift from universal noise mitigation toward structure- and noise-aware optimization strategies.