Driven by carbon neutrality goals, Integrated Energy Systems (IES) face the coupled challenges of high-dimensional nonlinear forecasting and combinatorial explosion in scheduling. Classical deep learning and optimization methods struggle with complex nonlinearities, large discrete spaces, and tightly coupled constraints. Meanwhile, advances in superconducting and trapped-ion hardware—scaling from tens to hundreds, and even approaching thousand qubits—together with error mitigation and hybrid quantum–classical architectures, have made the practical integration of quantum methods increasingly feasible. However, existing studies remain fragmented and task-specific, lacking a unified framework and an engineering-oriented assessment. This paper proposes a quantum-enhanced forecasting-scheduling pipeline that decomposes IES workflows into modular layers and systematically positions quantum methods within the closed loop. We show that, under Noisy Intermediate-Scale Quantum (NISQ) conditions, quantum approaches are most effective as plug-in computational enhancers, improving high-dimensional representation in forecasting and accelerating discrete optimization in scheduling. Based on this insight, we derive task-oriented method selection strategies and practical deployment guidelines, including a decision tree for layer selection, IES-specific penalty tuning, and fallback mechanisms for constraint violation. The proposed framework clarifies applicability boundaries and provides a structured pathway for transitioning quantum-enhanced IES from theoretical studies to engineering practice.
We study collaborative training of personalized quantum classifiers in a label-private setting in which each user privatizes labels locally via randomized response before upload and may choose an individual privacy budget. User-specific variational quantum models are trained on disjoint datasets, and their predictions are combined through voting-based or learned aggregation rules. Under the strict privacy-consistent regime, in which every downstream supervision signal is itself privatized before use, the protected-label stream inherits record-level ε _i -local differential privacy and the ensemble-level guarantee is bounded by max _i ε _i under parallel composition across disjoint users. A learned aggregation module may also be calibrated on a small clean validation set as an optional extension outside that strict formal privacy scope. Empirically, the method is competitive under strict local privacy, with the preferred aggregation rule depending on the supervision regime: voting is the most stable choice when the pipeline remains privacy-consistent end to end, whereas learned aggregation is most effective when reliable clean calibration labels are available. Additional experiments on multiclass tasks, partial participation, scalability, simulated noise, and a replicated IBM Quantum pilot study further characterize the operating envelope of the framework.
Quantum computing has emerged as a promising paradigm for addressing computational tasks intractable for classical systems, leveraging quantum mechanical principles such as superposition and entanglement to efficiently explore high-dimensional solution spaces. In recent years, hybrid quantum-classical approaches have gained increasing attention as a means to exploit the representational power of quantum systems while preserving the practicality of classical machine learning frameworks. This paper presents a study on quantum-enhanced model compression via feature-based knowledge distillation, in which a large classical neural network (the teacher) transfers knowledge to a substantially smaller hybrid student model. The proposed hybrid quantum-classical student architecture incorporates a parameterized quantum circuit (PQC) implemented in Qiskit, employing a linear R_y angle-encoding feature map followed by an EfficientSU2 variational ansatz, with features bounded through a tanh (· )×π/2 scaling to ensure stable gradient propagation. Gradient computation is performed via Reverse-Mode (adjoint) differentiation through the StatevectorEstimator primitive, which provides exact, shot-noise-free gradients at the cost of a single forward–backward pass per optimization step. The hybrid student is compared against a parameter-matched purely classical student under strict 1:1 parameter parity across a six-task binary benchmark spanning MNIST and Fashion-MNIST, as well as a four-class multi-class benchmark, totaling eleven classification experiments. A hardware validation experiment is additionally conducted on the IQM Garnet 20-qubit quantum processor using the Parameter-Shift Rule. The results indicate that the hybrid student outperforms its classical counterpart in five of six binary tasks (83
Artificial intelligence (AI) is transforming pharmaceutical science by enhancing efficiency and scalability across drug discovery, precision medicine, and clinical development. Unlike traditional methods hampered by high costs and low success rates, AI-driven approaches like machine learning enable faster target identification, drug repurposing, and trial prediction. Integrating multi-omics data supports data-driven decisions, while emerging tools, such as digital twins, large language models (LLMs), and quantum computing expand analytical capabilities. However, existing literature often treats classical AI, generative models, and quantum approaches separately, limiting insight into their comparative strengths. This PRISMA-based review synthesizes these domains, highlighting challenges in data quality, interpretability, ethics, and regulation. Findings show growing convergence between AI and precision therapeutics, but real-world application demands robust validation, collaboration, explainable AI, and global standards. The review offers a structured view of current advances, persistent gaps, and future directions in AI-enabled pharmaceutical innovation.
Quaternion-valued neural networks and variational quantum circuits (VQCs) both derive local transformations from SU(2) geometry, yet their relative performance on classical supervised learning tasks remains poorly understood. We present a controlled comparison in which real-valued, quaternion-valued, and quantum classification heads operate on identical frozen feature representations across MNIST, FashionMNIST, and CIFAR-10. CIFAR-10 experiments include both a learned 16-dimensional bottleneck and frozen ImageNet-pretrained ResNet18 features to separate architectural effects from representation quality. Quaternion classifiers consistently match or closely approach real-valued baselines while substantially outperforming shallow VQCs. On MNIST and FashionMNIST, quaternion networks achieve near-equivalence with real-valued multilayer perceptrons, whereas product-state VQCs exhibit lower accuracy and substantially higher computational cost. On CIFAR-10, quaternion networks retain 94–97 χ^2=12.796 , p = 0.0051 , n = 5 ), with post-hoc Wilcoxon signed-rank tests yielding large effect sizes ( d > 5 ) for all QuatNet vs. quantum comparisons on MNIST. For FashionMNIST and CIFAR-10, large effect sizes ( d > 2.0 ) serve as the primary inferential statistic given n = 3 . These results indicate that, on classical vision benchmarks lacking intrinsic quantum structure, quaternion networks provide efficient and stable SU(2) alternatives to shallow variational quantum circuits. The findings suggest that shared local SU(2) geometry and shallow entanglement are insufficient, within the shallow circuit regime studied here, to confer practical quantum advantage on classical image-classification tasks. The conclusions are bounded to shallow, measurement-limited variational circuits operating on classical image-classification tasks without intrinsic quantum structure.
The Quantum Whale Optimization Algorithm (QWOA), a novel nature-inspired, quantum-based metaheuristic optimization algorithm, is proposed in this study. This algorithm inherits the social behaviour of humpback whales, specifically their unique hunting approach, from WOA and adopts/integrates the concept of local attractor in quantum mechanics. QWOA is compared with nine algorithms including existing WOA variants. A nonlinear production-inventory model is formulated, incorporating dynamic power-based demand and nonlinear holding costs. The demand for products is also influenced by nonlinear factors such as selling price, green level, and warranty period. Additionally, the model includes carbon emission reduction investment as part of its formulation. To evaluate the robustness of the proposed algorithm, twenty-three classical mathematical benchmarks and IEEE CEC 2022 benchmarks are solved and compared with QWOA. The optimization results demonstrate that QWOA is highly competitive when compared to state-of-the-art metaheuristic algorithms. For critical comparison over the chosen metaheuristics, we used both parametric (ANOVA) and non-parametric (Kruskal–Wallis) tests. First order sensitivity analysis is performed for the inventory problem in order to make fruitful conclusion over the inventory model.
To address the challenges of poor adaptability in variational quantum algorithm ansatz templates and high noise sensitivity in deep structures on NISQ devices, an adaptive ansatz construction and optimization method is proposed. This approach dynamically adjusts circuit structures and optimizes performance according to varying datasets and task requirements. A full binary tree is used to represent the predefined ansatz structure, and a simulated annealing algorithm guides the qubit merging path to facilitate construction. By integrating hierarchical ansatz training with multi-objective particle swarm optimization, the method simultaneously optimizes expressibility, entanglement capability, and circuit depth. This ensures circuit depth compression while balancing these performance metrics, enabling task-oriented ansatz design and circuit optimization. On three binary classification tasks from the Iris dataset, the method achieved average accuracies of 98.3
Instance selection algorithms for supervised classification and regression tasks in machine learning aim to reduce the size of training datasets whilst maintaining the performance of models by selecting the most important, high-fidelity training labels from a large pool of instances in the original datasets. Traditional instance selection methods such as cluster-based techniques demonstrate enhanced generalization capabilities on fewer data points compared with random sampling approaches, but often suffer from high computational costs, sensitivity to hyperparameters and non-deterministic time complexities. This paper proposes QReduce which is a novel instance selection algorithm which models datasets as N-dimensional graphs with each data point corresponding to a vertex and edges constructed between points in a k-nearest neighbour manner. This graph is subjected to quantum annealing to solve the minimum vertex cover or maximum cut graph combinatorial optimization problems and yield a subgraph representing the reduced subset of instances selected for the reduced dataset. Solving such NP-hard problems are intractable for exact classical solvers on account of combinatorial explosion of the problem’s search space, and quantum annealing was shown to outperform other approximate metaheuristic techniques such as simulated annealing, justifying it as a strong metaheuristic optimization model. QReduce enables recursive reduction of the dataset and has the benefits of less sensitivity to hyperparameters as well as a deterministic time complexity, enabling efficient and effective reduction in storage and compute requirements for training complex machine learning models in the age of green and sustainable AI. Experimental results demonstrate the efficacy of QReduce, outperforming existing instance selection techniques and having the highest average percentage data reduction to percentage accuracy reduction ratios on datasets for both classification and regression tasks, thereby making it a best-in-class instance selection algorithm for selecting high-fidelity training labels in supervised machine learning tasks.
Recently, image retrieval has become an emerging field, and it plays a versatile role in industries such as marketing and design, health care, security, and entertainment. Traditional image retrieval systems have challenges with high-dimensional feature spaces, which lead to inefficiency in processing and retrieval times. Moreover, as image volume increases, maintaining performance and speed becomes increasingly challenging, particularly in real-time applications. Hence, the proposed study presents an advanced Content-Based Image Retrieval (CBIR) system that integrates Quantum Convolutional Neural Networks (QCNN) with Deep Reinforcement Learning (DRL) to enhance retrieval accuracy and adaptability, addressing challenges posed by high-dimensional feature spaces in traditional systems. The framework utilises quantum-inspired feature extraction alongside classical CNNs to effectively capture complex image representations. Experiments performed on diverse datasets (Caltech-101, CIFAR-10, FTVL) demonstrate that the proposed system achieves exceptional mean Average Precision (mAP) scores: 1.0 for 9 out of 10 classes in CIFAR-10 (with 0.99 for the remaining class), 0.972 on FTVL, and 0.905 on Caltech-101, significantly outperforming traditional models such as FDenseNet (0.958 on FTVL and 0.891 on Caltech-101). In comparative analysis across different bit lengths, the proposed model achieves mAP scores of 0.9754 (16-bit), 0.9858 (32-bit), 0.9931 (48-bit), and 0.9986 (64-bit), surpassing state-of-the-art methods. Ablation studies confirm that the combination of Quantum CNN with Synergistic Policy Improvement Meta-Optimization (SPIMO) and dynamic feature weighting attains the highest mAP of 1.0, validating the effectiveness of the proposed approach. The system employs cosine similarity for image comparison, and the incorporation of feedback mechanisms facilitates continuous learning, enhancing robustness in real-world applications. To support reproducibility, the source code and implementation details will be made publicly available upon acceptance of the manuscript and are available from the corresponding author upon reasonable request.
Hybrid classical-quantum neural network (HCQNN) models have recently emerged as a powerful approach for classification problems, due to their strong computational representation capabilities coming along with the integration of neural networks and quantum computing. However, these models are not free from the fundamental issue of poor separation between the in-distribution (ID) and Out-Of-Distribution (OOD) samples in the classification output space, which is observed in classical neural network classifiers too. Despite this, to the best of our knowledge, there is no current work that systematically studies the OOD detection problem in the context of the HCQNN models. Towards this end, we benchmark the existing approaches for OOD detection in the classical neural network literature domain on the HCQNN classifier models using the standard datasets and metrics and note their limitations. Thereby we propose a novel strategy suitable for OOD prediction in the HCQNN classifier models using the representational properties of the quantum features’ space. Particularly, we find subspaces within the feature space based on their categorical label information, and make use of a metric called the fidelity score that is extensively used in the quantum computing literature for measuring the similarity between the quantum states. Finally, we substantiate our claims on the efficacy of the fidelity score for OOD detection by demonstrating empirically that we can separate the ID from OOD samples across many standard benchmarks using the class-wise boundary defined characterized by the fidelity scores.
Semidefinite programming (SDP) serves a dual role in quantum cryptography: it is both the primary mathematical tool for certifying the security of quantum key distribution (QKD) protocols and, simultaneously, a computational bottleneck that limits the scale of protocols amenable to rigorous analysis. From device-independent security proofs using the Navascués-Pironio-Acín (NPA) hierarchy to finite-key rate calculations, SDPs provide rigorous mathematical frameworks for certifying quantum cryptographic security. However, as quantum protocols become more sophisticated–incorporating higher-dimensional systems, multiple measurement settings, or device-independent assumptions–the computational complexity of solving these SDPs classically becomes a significant bottleneck, scaling as O(n^3.5) for interior-point methods where n denotes the SDP dimension. Concurrently, quantum algorithms for SDP solving have emerged, offering polynomial speedups over classical methods. The Brandão–Svore algorithm and its refinements achieve Õ(√(mn)) query complexity versus Õ(mn) classically, a quadratic speedup that is near-optimal as established by quantum lower bounds. This raises a fundamentally self-referential question: Can quantum computers accelerate the security analysis of quantum cryptographic protocols? This review examines the dual role of SDPs in quantum cryptography, surveys SDP-based security proofs in QKD, reviews recent advances in quantum SDP algorithms (from fault-tolerant approaches to NISQ-era variational methods), and analyzes the theoretical and practical implications of using quantum computation to bootstrap quantum cryptographic security. We identify the practical quantum advantage threshold at problem sizes n · m > 10^12 , corresponding to high-dimensional QKD with d > 50 , device-independent protocols with more than five measurement settings at NPA level ≥ 3 , or quantum networks with N ≥ 5 nodes. We further establish that untrusted quantum SDP solving can be used safely under classical solution verification, preserving information-theoretic security guarantees. We identify this computational loop as a critical open question for the future of quantum cryptography and outline pathways toward practical quantum-accelerated security analysis.
Distributed Denial-of-Service (DDoS) attacks remain one of the most critical threats to modern cybersecurity. While machine learning techniques have proven effective for detection, classical approaches struggle with the growing complexity and scale of these attacks. Quantum computing, particularly quantum kernel methods, offers a promising alternative; however, the current state of the art faces a major challenge: vanishing similarity, which severely limits model expressiveness in high-dimensional spaces. This work introduces a novel quantum kernel inspired by multiple kernel learning, designed to mitigate vanishing similarity by constructing kernels in reduced-dimensional subspaces and combining them through averaging. The methodology is validated on the Canadian Institute for Cybersecurity dataset (NTP-based DDoS attacks). The proposed kernel effectively preserves classification capability in high-dimensional feature spaces, paving the way for practical applications of quantum kernels.
Behavioral biometric authentication from keystroke and smartphone-based gait signals is inherently a genuine-only learning problem under session variability and bounded neuromotor adaptation. We formulate this setting as one-class reconstruction in a fixed descriptor space and propose a Quantum–Classical Adaptive Autoencoding (QCAA) framework for modeling user-consistent behavioral structure. The framework integrates deterministic time-series feature construction, QUBO-based structured sparsification for circuit-width regulation, shallow variational quantum embedding, and measurement-induced reconstruction within a unified hybrid architecture. Reconstruction is derived directly from observable statistics, enabling anomaly scoring in the original feature domain while incorporating nonlinear Hilbert-space interactions. A controlled five-session dataset comprising fixed-text typing, free-text typing, and smartphone-based gait signals from 100 participants was constructed to evaluate longitudinal behavioral stability. Under session-partitioned one-class evaluation in the primary noise-aware simulation setting, the optimized QCAA achieves equal error rates as low as 2.60
Quantum information science is rapidly advancing toward practical implementations, with photonic technologies emerging as a central enabler due to their unique advantages in scalability, low-loss transmission, and room-temperature operation. This review presents a comprehensive analysis of recent developments that bridge quantum light and computation, covering material platforms, device architectures, and application domains in photonic quantum computing. We systematically examine integrated photonic circuits, single-photon sources, quantum memories, and photonic logic gates, highlighting their roles in realizing robust and efficient quantum information processing. Emphasis is placed on state-of-the-art experimental demonstrations and theoretical frameworks that integrate quantum photonics with emerging computational paradigms, including machine learning and hybrid quantum–classical systems. The review further addresses challenges such as loss management, photon indistinguishability, and large-scale integration, while discussing potential breakthroughs through novel material systems, topological photonics, and quantum networking protocols. By synthesizing recent progress and identifying key bottlenecks, this work provides a forward-looking perspective on how photonic technologies can drive the next generation of quantum computing and communication systems.
Quantum neural network classifiers have attracted significant attention due to their remarkable performance in classification tasks. However, when dealing with high-dimensional data, these classifiers are often susceptible to misclassification caused by carefully crafted perturbations, a type of attack known as adversarial attacks. To address this issue, we propose a quantum adversarial attack algorithm—Q-MIFGSM. This algorithm generates perturbations by analyzing the gradient information of the input data and combining it with momentum, effectively disrupting the performance of a trained Quantum Neural Network classifier. Compared to existing quantum attack algorithms baselines (Q-FGSM and Q-BIM), experimental results explicitly show that Q-MIFGSM demonstrates superior attack efficacy and faster adversarial sample learning on both Fashion-MNIST and MNIST datasets. Meanwhile, noisy Q-MIFGSM achieves enhanced attack efficiency instead of performance degradation under five types of quantum noise. This study not only reveals the vulnerabilities of quantum classifiers but also contributes to the understanding of quantum adversarial attacks and adversarial training.
The smart grid’s extensive digitalization and integration of renewables have expanded its cyber-physical attack surface, exposing critical systems to ransomware-style intrusions and nation-state shutdowns. Recent NIST and regulatory assessments reveal that process sensors–devices that convert physical parameters into grid control measurements–lack fundamental cybersecurity measures, including authentication and cyber forensic capabilities. In contrast, conventional monitoring only validates protocol patterns, rather than the authenticity of physical measurements themselves. We propose a triple-layer quantum defense framework addressing this security gap by combining quantum communication protocols with quantum sensing and machine learning. The first (lightweight) layer employs quantum teleportation with a Shared Entanglement (SE) security model, where nodes encode measurements as quantum states and transmit them with inherent tamper detection, providing a communication-layer integrity check. Complementing this, the second (heavyweight) layer operates at the node level: a hybrid architecture merging legacy classical current transformers with quantum sensors using Greenberger-Horne-Zeilinger (GHZ) entangled states. Through distributed entangled correlations across geographically dispersed sensors, this layer inherently detects physical-layer tampering. For sophisticated, coordinated attacks that bypass the checks in earlier layers, we introduce a third last-resort (post-compromise) layer: a quantum key distribution (QKD) or post-quantum cryptography (PQC) secured transfer with a centralized machine learning algorithm operating on raw Automatic Generation Control (AGC) time series, identifying anomalous patterns and achieving 98.7
This paper presents the StateQ framework, a novel approach in quantum computing that emphasizes state-centric programming over traditional quantum circuit construction. At its core is the StateQ language and its accompanying compiler, designed for intuitive understanding and implementation of quantum algorithms. The framework also includes the Quantum Circuit Transpiler and Low-Level Quantum Assembly (LLQASM), enhancing circuit adaptability across different hardware. Additionally, the Quantum Intermediate Virtual Machine (QIVM) and Quantum Interface Layer (QIL) ensure smooth execution and compatibility of StateQ programs on various platforms. This paper unfolds the essence of StateQ, spotlighting its design and core features, heralding a structured and intuitive framework for quantum programming.
This research presents a comprehensive framework for secure healthcare prediction by integrating quantum-inspired heuristic algorithms with blockchain technology. The framework comprises three core components: privacy-preserving feature extraction, robust data classification, and secure data management. First, we leverage Privacy-Preserving Generative Adversarial Networks (PPGANs) to extract synthetic features from healthcare data. These features retain the statistical properties of the original data while safeguarding sensitive information, enabling practical analysis without compromising patient privacy. Next, we introduce an advanced classification methodology that combines Federated Learning with Quantum-inspired Particle Swarm Optimization (QPSO) for hyperparameter optimization. This approach enhances the accuracy of classification models while preserving data privacy across distributed networks, ensuring reliable and robust predictions in healthcare applications. Finally, we explore integrating blockchain technology to secure data storage and management. The proposed model achieved an accuracy of 99.01
