You Only Measure Once: On Designing Single-Shot Quantum Machine Learning Models

2025-09-24

AI Summary

Quantum machine learning (QML) inference typically requires many measurement shots to estimate observable expectation values, incurring high hardware costs and latency—particularly prohibitive on real-world quantum platforms where shots are billed individually. To address this, we propose Yomo, the first framework enabling high-accuracy QML inference with a single shot. Yomo replaces conventional Pauli expectation estimation with a probabilistic aggregation mechanism and introduces a sharpened loss function alongside noise-robust training strategies, markedly improving prediction reliability in the ultra-low-shot regime. In simulated experiments on MNIST and CIFAR-10 under depolarizing noise, Yomo achieves performance comparable to multi-shot baseline models using only one shot, consistently outperforming state-of-the-art alternatives in accuracy. This work breaks the scalability dependence of QML on shot count, establishing an efficient, low-cost inference paradigm toward practical quantum machine learning.

Technology Category

• Machine Learning: Quantum Machine Learning
• Search and Optimization: Learning to Search
• Humans and AI: Human-in-the-loop Machine Learning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Quantum machine learning (QML) models conventionally rely on repeated measurements (shots) of observables to obtain reliable predictions. This dependence on large shot budgets leads to high inference cost and time overhead, which is particularly problematic as quantum hardware access is typically priced proportionally to the number of shots. In this work we propose You Only Measure Once (Yomo), a simple yet effective design that achieves accurate inference with dramatically fewer measurements, down to the single-shot regime. Yomo replaces Pauli expectation-value outputs with a probability aggregation mechanism and introduces loss functions that encourage sharp predictions. Our theoretical analysis shows that Yomo avoids the shot-scaling limitations inherent to expectation-based models, and our experiments on MNIST and CIFAR-10 confirm that Yomo consistently outperforms baselines across different shot budgets and under simulations with depolarizing channels. By enabling accurate single-shot inference, Yomo substantially reduces the financial and computational costs of deploying QML, thereby lowering the barrier to practical adoption of QML.

Problem

• Reduces quantum machine learning inference costs by minimizing measurement shots
• Replaces Pauli expectation outputs with probability aggregation for efficiency
• Enables accurate single-shot predictions to lower deployment barriers

Innovation

• Probability aggregation replaces Pauli expectation-value outputs
• Loss functions encourage sharp predictions for accuracy
• Single-shot inference reduces quantum hardware measurement costs