How Much Reconstruction Does Quantum Machine Learning Need? Late Fusion of Independently Trained Quantum Subcircuits

2026-08-06

AI Summary

In quantum machine learning, reconstructing outputs after circuit cutting incurs an exponential overhead in classical sampling, posing a major bottleneck. This work proposes a late-fusion strategy that partitions large-scale quantum neural networks into independent subcircuits, which are trained and measured separately, and then combines their outputs via a lightweight classical decision head through linear fusion at the output layer—thereby circumventing costly full-state reconstruction. The approach innovatively introduces a tunable "quantumness" parameter $Q$ and a cut-entanglement diagnostic tool to adaptively assess the necessity of reconstruction. Notably, it is the first to integrate the late-fusion mechanism from multimodal learning into quantum machine learning. Experiments on synthetic and standard datasets demonstrate that the method achieves accuracy comparable to full reconstruction (error ≤ 0.04) with exponentially lower computational cost and significantly enhanced robustness against sampling and device noise.

Abstract

Circuit cutting lets a large quantum neural network (QNN) run as independent subcircuits on small devices, but rebuilding its outputs by reconstruction carries a classical sampling overhead exponential in the number of cuts - the dominant runtime cost in prior work. We ask whether, for machine-learning tasks, this step is necessary, and replace it with late fusion: each subcircuit is trained and measured independently, and a small classical head combines their outputs - a linear-cost, decision-level combination borrowed from multimodal learning. To characterize the trade-off we introduce a quantumness dial $Q$, a tunable reconstruction budget interpolating from pure fusion to full reconstruction, and a cut-entanglement diagnostic that indicates how much reconstruction a task needs (Spearman $ρ=0.59$ over $104$ runs). Across synthetic and standard datasets, independently trained late fusion matches full reconstruction accuracy within $0.04$ at every point of the controlled sweep and on every classical benchmark, at exponentially lower cost; it is also markedly more robust to shot and device noise. Controlled entangled-data experiments locate the boundary where fusion must fail. We do not claim advantage over classical machine learning - consistent with recent benchmarking, quantum offers no accuracy edge on these datasets. Late fusion is thus an efficient, noise-robust, self-characterizing alternative to reconstruction for circuit-cutting QML.

Problem

• quantum machine learning
• circuit cutting
• reconstruction overhead
• late fusion
• quantum neural networks

Innovation

• late fusion
• circuit cutting
• quantum machine learning
• reconstruction overhead
• quantumness dial