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.
arXiv:2608. 05595v1 Announce Type: cross 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.
arXiv:2604. 06135v2 Announce Type: replace-cross Abstract: Efficient data loading remains a bottleneck for near-term quantum machine learning.
arXiv:2510. 03389v2 Announce Type: replace-cross Abstract: Current quantum computers require algorithms that use limited resources economically.
arXiv:2606. 26312v1 Announce Type: cross Abstract: Autoencoders transformed classical machine learning by solving the curse of dimensionality, enabling principled weight initialization and learning compact, structured representations.
Quantum MeanFlow (QMF) is a new quantum generative sampling method that enables single‑step sample generation by learning an average velocity field over a time interval, unlike the multi‑step quantum flow matching (QFM) which requires sequential integration of an ordinary differential equation. Using parameterized quantum circuits, the authors benchmark QMF and QFM on the MNIST dataset, finding that QMF produces lower image quality than multi‑step QFM but outperforms single‑step QFM at every shot count. Both models were executed on IBM quantum computers, and best‑of‑N rejection sampling mitigates device noise without circuit modification, demonstrating QMF’s practicality for efficient single‑step quantum generative sampling.
arXiv:2607. 00365v1 Announce Type: cross Abstract: Artificial intelligence (AI) and quantum information (QI) are rapidly co-evolving.
arXiv:2508. 19437v2 Announce Type: replace-cross Abstract: The importance of analyzing nontrivial datasets when testing quantum machine learning (QML) models is becoming increasingly prominent in literature, yet a cohesive framework for understanding dataset characteristics remains elusive.
Bridge of Ψ's (BOPS) is a generative model that learns to transform quantum circuits into equivalent, optimized versions using Schr"odinger bridges and a custom denoiser architecture. Trained on data engineered to challenge existing optimizers, BOPS achieves a 2.46× reduction in gate count and a 2.45× reduction in depth on 8‑qubit, 64‑depth Clifford+$T$ circuits, outperforming nine baseline optimizers. This work demonstrates the first successful application of generative machine learning to quantum circuit optimization, expanding the quantum compilation stack with learned techniques.
The paper introduces Conformalized Quantum DeepONet Ensembles, a framework that combines Quantum Orthogonal Neural Networks (QOrthoNNs) with split conformal calibration to address the quadratic cost of dense neural layers and unreliable uncertainty quantification in operator learning. It demonstrates that QOrthoNNs achieve ×O(n) hidden‑layer running‑time scaling, improving on the ×O(n^2) cost of classical dense layers, while the ensemble approach provides a finite‑sample, distribution‑free lower bound on coverage for new input‑output function pairs. Experiments on synthetic benchmarks and real‑world power‑system dynamics confirm accurate predictions and empirical coverage close to the target under both ideal and noisy quantum simulations.
The paper investigates how postprocessing routines in quantum neural network software can inadvertently discard a large portion of valid measurement data when run on real quantum hardware. In a case study of Qiskit’s “SamplerQNN”, a filter that assumes virtual qubit space caused 85–99.6% of measurement shots to be lost on IBM backends, leading to unnormalised probability vectors, degraded inference accuracy (from 0.94 to 0.39), and a 22–27× compression of the training loss signal. The authors provide a layout‑based marginalisation fix that has been merged into the library to ensure forward‑compatibility with current and future hardware.
The paper introduces a hybrid quantum–classical regression framework that uses a lightweight classical embedding as a learnable geometric preconditioner to improve the conditioning of a downstream variational quantum circuit. It further incorporates a curriculum optimization protocol that gradually increases circuit depth and switches from SPSA-based exploration to Adam-based fine‑tuning. Experiments on PDE‑informed and standard regression datasets show that this approach consistently outperforms pure QNN baselines, yielding more stable convergence and reduced structured errors, especially in data‑limited regimes.
The paper investigates how postprocessing routines in quantum neural network software can cause significant data loss when run on large quantum hardware. In a case study of Qiskit’s “SamplerQNN”, a filter that assumes measurement bit‑strings are in virtual qubit space removed 85–99.6% of valid shots on IBM backends, leading to unnormalised probability vectors and distorted predictions. This loss caused inference accuracy to drop from 0.94 to 0.39 and compressed training loss signals by 22–27×, severely reducing optimizer sensitivity. The authors implemented a layout‑based marginalisation fix that was merged into the library to make “SamplerQNN” forward‑compatible with current and future hardware.