Automated visual inspection underpins quality assurance in intelligent manufacturing. However, the convolutional and transformer models used for this purpose can be costly to deploy at the network edge, where latency and energy budgets are tight. Variational quantum circuits (VQCs) are often advanced as a low-parameter alternative, the reasoning being that n qubits carry a state space of size 2ⁿ, so that comparatively few trainable angles might still describe an intricate decision surface. This selective narrative review asks whether that reasoning survives once resources are tallied from sensor input through to decision. A framework proposed that separates task-specific trainable parameters, stored non-trainable parameters, and fitted quantities, preprocessing compression, and quantum encoding reach, and apply it to representative hybrid pipelines. Three findings emerge. First, claimed quantum gains tend to arise in studies where the classical baseline differs from the quantum pipeline in parameter count, preprocessing applied, or the amount of tuning each received. A prominent controlled benchmark for small binary classification tasks generally favors classical models, although direct extrapolation of that result to industrial vision remains unproven. Second, the selected examples show few trainable angles within the circuit, but the representational burden is largely shifted onto classical preprocessing or frozen backbones. Third, scalability is constrained by both encoding reach and state preparation: a standard video clip can contain nearly 5 × 106 values, and generic amplitude loading may require millions of operations. These constraints are particularly relevant to video and sequential-image process monitoring. Six reporting requirements are proposed to make future comparisons reproducible.
Automated visual inspection underpins quality assurance in intelligent manufacturing. However, the convolutional and transformer models used for this purpose can be costly to deploy at the network edge, where latency and energy budgets are tight. Variational quantum circuits (VQCs) are often advanced as a low-parameter alternative, the reasoning being that n qubits carry a state space of size 2ⁿ, so that comparatively few trainable angles might still describe an intricate decision surface. This selective narrative review asks whether that reasoning survives once resources are tallied from sensor input through to decision. A framework proposed that separates task-specific trainable parameters, stored non-trainable parameters, and fitted quantities, preprocessing compression, and quantum encoding reach, and apply it to representative hybrid pipelines. Three findings emerge. First, claimed quantum gains tend to arise in studies where the classical baseline differs from the quantum pipeline in parameter count, preprocessing applied, or the amount of tuning each received. A prominent controlled benchmark for small binary classification tasks generally favors classical models, although direct extrapolation of that result to industrial vision remains unproven. Second, the selected examples show few trainable angles within the circuit, but the representational burden is largely shifted onto classical preprocessing or frozen backbones. Third, scalability is constrained by both encoding reach and state preparation: a standard video clip can contain nearly 5 × 106 values, and generic amplitude loading may require millions of operations. These constraints are particularly relevant to video and sequential-image process monitoring. Six reporting requirements are proposed to make future comparisons reproducible.
Abstract Quantum information processing is gaining popularity in the fields of machine learning and image processing because of its advantages. Quantum convolution is an interesting topic in this field, and studies on this topic can be divided into value-based and angle-based methods. Although quantum convolution studies on angle-based or variational quantum circuits (VQCs) is called convolution, the circuits work differently from classical convolution. In this study, contrary to the literature, the VQC was trained to imitate classical convolution. The differential evolution algorithm (DEA) was used to optimize the VQCs. The proposed method requires as many qubits as the filter size (N× N). The generated circuits contain N× N× 4 quantum gates and N× N × 3 trainable parameters. The generated circuits were tested in Python environment using Cirq simulator. The Cifar10 and MNIST datasets are used as examples. For 2 × 2 filters with different weights, the convolution was successfully modeled with a mean squared error of less than 0.001. In general, the proposed method imitates classic convolution within ± 5 % tolerance. In conclusion, VQCs that imitate classical convolution with fewer qubits and quantum gates than value-based methods were obtained.
Quantum machine learning offers a compelling representational perspective: angle-encoded states inhabit Hilbert spaces in which periodic similarities and interactions can be expressed naturally. Realizing this perspective for visual recognition remains difficult, however, because present quantum neural networks are constrained by limited qubit counts, costly circuit simulation and measurement, noise, and unstable optimization on noisy intermediate-scale quantum devices. We investigate whether useful structural ideas from quantum models can instead be realized as scalable classical Transformer operations. We introduce QiT, a Quantum-inspired Transformer for vision tasks with three components: (i) angle-inspired encoding that maps image tokens to learned trigonometric Hilbert-space features analogous to quantum rotation-based state encoding; (ii) self-attention over these periodic features, inducing a classical cosine kernel approximated to quantum fidelity kernels; and (iii) gated multiplicative emulation, a trainable classical surrogate for interaction terms found in variational circuits. All components are differentiable tensor operations, so QiT claims neither quantum computation nor quantum speedup and retains the O(N2D) attention complexity of a standard Vision Transformer. Across image-classification benchmarks, QiT is competitive with a matched classical Transformer while avoiding the severe runtime cost observed for a small simulated quantum Transformer. QiT-B reaches 78.3\% ImageNet-1K top-1 accuracy with 45.7M parameters and 11.5 GFLOPs. These results position QiT as a scalable baseline for isolating and evaluating quantum-motivated inductive biases in visual recognition.
This report introduces a novel class of reasoning architectures, termed Quantum Circuit Reasoning Models (QCRM), which extend the concept of Variational Quantum Circuits (VQC) from energy minimization and classification tasks to structured logical inference and reasoning. We posit that fundamental quantum mechanical operations, superposition, entanglement, interference, and measurement, naturally map to essential reasoning primitives such as hypothesis branching, constraint propagation, consistency enforcement, and decision making. The resulting framework combines quantum-inspired computation with differentiable optimization, enabling reasoning to emerge as a process of amplitude evolution and interference-driven selection of self-consistent states. We develop the mathematical foundation of QCRM, define its parameterized circuit architecture, and show how logical rules can be encoded as unitary transformations over proposition-qubit states. We further formalize a training objective grounded in classical gradient descent over circuit parameters and discuss simulation-based implementations on classical hardware. Finally, we propose the Quantum Reasoning Layer (QRL) as a differentiable hybrid component for composable reasoning models applicable to scientific, biomedical, and chemical inference domains.
Variational quantum circuits (VQCs) face an expressivity-trainability dilemma and scalability challenges. We propose Multi-Layer Fully-Connected Variational Quantum Circuits (FC-VQC), a general-purpose quantum machine learning framework that connects local VQC blocks through measurement, deterministic parameter-free routing, and re-encoding. All trainable model parameters reside within the quantum blocks, without trainable classical neural components. We study fully connected, sliding-window, and parallel block mixing and analyze computational costs, conditional error propagation, and block information exchange. Using classical simulation, we evaluate tabular regression and classification, with our main comparison addressing spatio-temporal function approximation for the Black--Scholes, Burgers, and time-dependent oscillatory PDEs with up to 144 spatial dimensions. Comparisons include neural-network, explicit angle-feature, tensor-network, and gradient-boosted-tree baselines. The results show that FC-VQC achieves the lowest trajectory relative MAE for all PDEs at the higher dimensions d∈{81,144}. Gradient-dynamics and depolarizing-noise experiments provide complementary empirical diagnostics of trainability and preliminary noise sensitivity.
Abstract Variational quantum circuits (VQCs) are a central paradigm for quantum machine learning, yet their practical deployment is limited by poor trainability, sensitivity to noise, and the lack of input-adaptive parameterization. In this work, we introduce TensorDual-VQC, a tensor-structured operator-learning framework that models variational quantum computation as an input-conditioned mapping that jointly governs quantum state encoding and circuit parameter generation. The proposed approach employs a shared low-rank Tensor-Train (TT) backbone with dual output heads, inducing a coupled representation that constrains the hypothesis space to a structured low-dimensional manifold. Under standard smoothness and low-rank assumptions, we show that the approximation behavior is governed by the TT-rank rather than the ambient parameter dimension, and that shared latent structure can enable more compact representations than decoupled parameterizations in structured regimes. We further analyze trainability through the induced Neural Tangent Kernel, showing that the multi-branch gradient structure enriches the span of gradient features and can improve kernel conditioning under mild non-degeneracy conditions. In addition, we relate robustness to quantum noise to the intrinsic effective dimension of the tensor parameter manifold, providing a first-order scaling characterization of output variance under stochastic perturbations. We evaluate TensorDual-VQC on quantum dot classification, quantum simulation, and combinatorial optimization tasks under both simulated noise and real quantum hardware. Across all settings, the proposed framework achieves higher accuracy, faster convergence, and greater stability than standard VQCs and hypernetwork-based baselines when quantum resources are matched. These results suggest that structured operator learning with shared low-rank representations provides a principled and practical approach to scalable, noise-resilient quantum learning in the NISQ regime.
Quantum Transfer Learning (QTL) offers a promising approach for visual quantum machine learning under near-term constraints, where limited qubit counts, shallow circuit depths, and costly hybrid optimization restrict end-to-end quantum training. In this setting, pretrained classical backbones can extract high-level visual features, while compact quantum modules operate as trainable classification heads. However, existing QTL results are difficult to compare because they often differ in datasets, preprocessing, backbone settings, qubit budgets, circuit designs, optimization choices, and reporting protocols. This work presents a controlled benchmarking methodology for evaluating representative QTL methods under a unified transfer-learning pipeline. The benchmark compares DQN-QTL, QPIE-QTL, AE-CQTL, PVCQTL, and ED-QTL under shared preprocessing rules, frozen-backbone settings, training conditions, and reporting metrics. The evaluation focuses on Fashion-MNIST and Hymenoptera Ants vs. Bees as the two main datasets, while CIFAR-10 is used to provide additional configuration-level evidence on a harder natural-image task. Beyond predictive performance, the benchmark analyzes circuit size, trainable parameters, quantum parameters, training time, and architectural sensitivity to qubit count and circuit depth. The results show that no single QTL family dominates across all settings: performance depends on the dataset, encoding strategy, circuit design, and computational cost. These findings highlight the need for resource-aware QTL evaluation and provide guidance for selecting hybrid quantum-classical transfer models under near-term resource constraints.