Quantum machine learning (QML) provides parameter-efficient substitutes for traditional deep networks in medical imaging settings with constrained computational resources. However, deployment on noisy intermediate-scale quantum (NISQ) devices is difficult due to manual architecture design and decoherence constraints, particularly when circuit depth is limited. We introduce D-HAQNAS (Differentiable HardwareAware Quantum Neural Architecture Search), a differentiable hardware-aware framework that uses end-to-end differentiable optimization to find energy-efficient quantum topologies. Our method outperforms evolutionary methods in architecture search by up to$15 \times$. Three essential elements are integrated into the suggested framework: (1) a differentiable super-circuit with sigmoid-gated architectural parameters that permits gradient-based single-qubit gate topology search; (2) hardware-aware penalty terms that constrain gate count and circuit depth during training; and (3) weighted residual fusion to stabilize hybrid classical-quantum gradient flow. D-HAQNAS achieves 96.21% accuracy on PneumoniaMNIST, which is statistically comparable to classical baselines but with 78.6% fewer trainable parameters and a 31.5% decrease in the number of gates (from 54 to 37). D-HAQNAS shows statistically significant improvements on the severely data-limited RetinaMNIST dataset (1,080 samples, 5-class diabetic retinopathy classification) with an accuracy of 57.50% compared to 53.33% for classical models (+4.17 percentage point absolute gain, paired$t$-test$p=0.031)$. These findings suggest that parameter-efficient hybrid quantum-classical networks are a viable strategy for clinical settings with limited data, especially those where parameter efficiency is crucial.
Quantum machine learning (QML) provides parameter-efficient substitutes for traditional deep networks in medical imaging settings with constrained computational resources. However, deployment on noisy intermediate-scale quantum (NISQ) devices is difficult due to manual architecture design and decoherence constraints, particularly when circuit depth is limited. We introduce D-HAQNAS (Differentiable HardwareAware Quantum Neural Architecture Search), a differentiable hardware-aware framework that uses end-to-end differentiable optimization to find energy-efficient quantum topologies. Our method outperforms evolutionary methods in architecture search by up to15×. Three essential elements are integrated into the suggested framework: (1) a differentiable super-circuit with sigmoid-gated architectural parameters that permits gradient-based single-qubit gate topology search; (2) hardware-aware penalty terms that constrain gate count and circuit depth during training; and (3) weighted residual fusion to stabilize hybrid classical-quantum gradient flow. D-HAQNAS achieves 96.21% accuracy on PneumoniaMNIST, which is statistically comparable to classical baselines but with 78.6% fewer trainable parameters and a 31.5% decrease in the number of gates (from 54 to 37). D-HAQNAS shows statistically significant improvements on the severely data-limited RetinaMNIST dataset (1,080 samples, 5-class diabetic retinopathy classification) with an accuracy of 57.50% compared to 53.33% for classical models (+4.17 percentage point absolute gain, pairedt-testp=0.031). These findings suggest that parameter-efficient hybrid quantum-classical networks are a viable strategy for clinical settings with limited data, especially those where parameter efficiency is crucial.
Hybrid quantum-classical neural networks (HQNNs) are emerging as a practical approach for quantum machine learning in the noisy intermediate-scale quantum (NISQ) era, as they combine classical learning components with parameterized quantum circuits in an end-to-end trainable framework. However, their performance and efficiency depend strongly on architectural choices such as data encoding, circuit structure, measurement design, and the coupling between classical and quantum modules. This makes manual design increasingly difficult, especially when hardware limitations and resource constraints must also be taken into account. In this paper, we study the foundations of HQNNs and neural architecture search (NAS), discuss how NAS extends to quantum and hybrid settings, and demonstrate FLOPs-aware search (where FLOPs serve as a proxy for computational complexity), as an important hardware-aware direction for building HQNNs that are not only accurate but also computationally efficient and practically deployable.
Deploying quantum machine learning algorithms on near-term quantum hardware requires circuits that respect device-specific gate sets, connectivity constraints, and noise characteristics. We present a hardware-aware Neural Architecture Search (NAS) approach for designing quantum feature maps that are natively executable on IBM quantum processors without transpilation overhead. Using genetic algorithms to evolve circuit architectures constrained to IBM Torino native gates (ECR, RZ, SX, X), we demonstrate that automated architecture search can discover quantum Support Vector Machine (QSVM) feature maps achieving competitive performance while guaranteeing hardware compatibility. Evaluated on the UCI Breast Cancer Wisconsin dataset, our hardware-aware NAS discovers a 12-gate circuit using exclusively IBM native gates (6 ECR, 3 SX, 3 RZ) that achieves 91.23 % accuracy on 10 qubits-matching unconstrained gate search while requiring zero transpilation. This represents a 27 percentage point improvement over hand-crafted quantum feature maps (64 % accuracy) and approaches the classical RBF SVM baseline (93 %). We show that removing architectural constraints (fixed RZ placement) within hardware-aware search yields 3.5 percentage point gains, and that 100 % native gate usage eliminates decomposition errors that plague universal gate compilations. Our work demonstrates that hardware-aware NAS makes quantum kernel methods practically deployable on current noisy intermediate-scale quantum (NISQ) devices, with circuit architectures ready for immediate execution without modification.
Designing quantum neural networks (QNNs) that are both accurate and deployable on NISQ hardware is challenging. Handcrafted ansatze must balance expressivity, trainability, and resource use, while limited qubits often necessitate circuit cutting. Existing quantum architecture search methods primarily optimize accuracy while only heuristically controlling quantum and mostly ignore the exponential overhead of circuit cutting. We introduce QNAS, a neural architecture search framework that unifies hardware aware evaluation, multi objective optimization, and cutting overhead awareness for hybrid quantum classical neural networks (HQNNs). QNAS trains a shared parameter SuperCircuit and uses NSGA-II to optimize three objectives jointly: (i) validation error, (ii) a runtime cost proxy measuring wall clock evaluation time, and (iii) the estimated number of subcircuits under a target qubit budget. QNAS evaluates candidate HQNNs under a few epochs of training and discovers clear Pareto fronts that reveal tradeoffs between accuracy, efficiency, and cutting overhead. Across MNIST, Fashion-MNIST, and Iris benchmarks, we observe that embedding type and CNOT mode selection significantly impact both accuracy and efficiency, with angle-y embedding and sparse entangling patterns outperforming other configurations on image datasets, and amplitude embedding excelling on tabular data (Iris). On MNIST, the best architecture achieves 97.16% test accuracy with a compact 8 qubit, 2 layer circuit; on the more challenging Fashion-MNIST, 87.38% with a 5 qubit, 2 layer circuit; and on Iris, 100% validation accuracy with a 4 qubit, 2 layer circuit. QNAS surfaces these design insights automatically during search, guiding practitioners toward architectures that balance accuracy, resource efficiency, and practical deployability on current hardware.
Although quantum computers will not necessarily transform ML in general, they hold immense potential for specific uses, such as dealing with high-dimensional data and complex correlations that cannot be effectively modeled using classical systems. It is, however, difficult to design efficient quantum circuits, especially with the existing Noisy Intermediate-Scale Quantum (NISQ) hardware that has constraints including decoherence, a limited number of qubits, and gate fidelity limitations. The authors propose an Adaptive Quantum Circuit Architecture Search (AQCAS) framework whose structure and parameters are optimized dynamically to improve the performance of the task-specific ML. The framework integrates quantum circuit encoding, a hybrid search strategy that is based on reinforcement learning and evolutionary algorithms, and a task-specific adaptivity mechanism. Benchmark datasets were used for experimental evaluations of both simulation using Qiskit Aer and classical optimization using the Python frameworks TensorFlow and PennyLane. Fixed and standard variational quantum circuits were compared with performance measurements like the classification accuracy, circuit depth, convergence speed, and computational efficiency. The results show that the accuracy of AQCAS can be raised to 94.1%, and the circuit depth is lowered at the same time. It converged rapidly and is computationally efficient. Ablation studies show that both hybrid search and adaptivity mechanisms help to enhance generalization and stability under NISQ constraints. The results demonstrate the potential of AQCAS as a practical approach to using adaptive quantum circuits in specialized ML applications, and lay the groundwork for future development of scaling to larger devices, implementing noise-aware optimization, and integrating with hybrid quantum-classical ML pipelines.
Current quantum neural networks suffer from extreme sensitivity to both adversarial perturbations and hardware noise, creating a significant barrier to real-world deployment. Existing robustness techniques typically sacrifice clean accuracy or require prohibitive computational resources. We propose a hybrid quantum-classical Differentiable Quantum Architecture Search (DQAS) framework that addresses these limitations by jointly optimizing circuit structure and robustness through gradient-based methods. Our approach enhances traditional DQAS with a lightweight Classical Noise Layer applied before quantum processing, enabling simultaneous optimization of gate selection and noise parameters. This design preserves the quantum circuit's integrity while introducing trainable perturbations that enhance robustness without compromising standard performance. Experimental validation on MNIST, FashionMNIST, and CIFAR datasets shows consistent improvements in both clean and adversarial accuracy compared to existing quantum architecture search methods. Under various attack scenarios, including Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), Basic Iterative Method (BIM), and Momentum Iterative Method (MIM), and under realistic quantum noise conditions, our hybrid framework maintains superior performance. Testing on actual quantum hardware confirms the practical viability of discovered architectures. These results demonstrate that strategic classical preprocessing combined with differentiable quantum architecture optimization can significantly enhance quantum neural network robustness while maintaining computational efficiency.
Abstract The arrival of noisy intermediate-scale quantum (NISQ) devices defines an era in which quantum utility is fundamentally constrained by finite coherence times, limited qubit counts, gate infidelities, restricted connectivity, and measurement noise. Within this regime, hybrid quantum-classical frameworks have emerged as the dominant operational paradigm, particularly for quantum machine learning (QML). This review identifies a fundamental trade-off between expressivity, trainability, and noise resilience that governs the performance and scalability of NISQ-era QML systems. To contextualize these challenges, we introduce the resource-constrained iterative routing framework, a hardware-aware conceptual architecture that integrates topology-aware classical compression and distributed quantum orchestration to maintain circuit depths within device-dependent limits while routing representations toward symmetry-constrained Hilbert subspaces to mitigate noise-induced barren plateaus. Furthermore, we shift the discussion of quantum advantage away from asymptotic fault-tolerant complexity toward the notion of operational NISQ advantage based on finite-resource scaling, trainability preservation, and sampling complexity. We evaluate the practical resource efficiency of variational quantum algorithms, quantum kernel methods, and quantum neural networks across superconducting, trapped-ion, neutral-atom, and photonic platforms. The review also examines reproducibility challenges, emphasizing benchmarking against optimized classical baselines, and discusses strategies for managing the sampling overhead of error mitigation alongside distributed execution and quantum circuit compilation. Ultimately, NISQ in Practice positions hybrid QML as a physically constrained computational paradigm while outlining a pathway from contemporary noisy workflows toward early fault-tolerant quantum architectures.