Early detection of plant leaf diseases is critical for minimizing crop losses and supporting precision agriculture. While Convolutional Neural Networks (CNNs) have demonstrated high accuracy in image-based diagnosis, conventional architectures may not optimally balance spatial localization and channel-wise feature refinement, particularly in multi-crop classification settings. This study proposes a redundancy-aware dual-attention architecture, termed ATSA-DenseNet, which integrates the spatial branch of the Convolutional Block Attention Module (CBAM-Spatial) with Efficient Channel Attention (ECA) within a DenseNet121 backbone. Unlike prior dual-attention frameworks that retain full CBAM and introduce channel-level redundancy, the proposed design isolates complementary spatial and channel mechanisms to improve representational efficiency without increasing computational complexity. The framework is evaluated on controlled multi-crop PlantVillage-derived datasets comprising tomato, potato, pepper, and maize. Across both 3-crop and 4-crop configuration, ATSA-DenseNet consistently outperforms baseline DenseNet121 and single-attention variants, achieving 99.94% accuracy and 0.9994 macro-F1 on the 4-crop setting while maintaining a lightweight footprint (6.96M parameters, 2.87G FLOPs). Grad-CAM visualizations indicate improved localization of disease-relevant regions compared to the baseline. While results are obtained under controlled imaging conditions, the findings demonstrate that redundancy-aware dual-attention enhances feature discrimination efficiency in multi-class agricultural classification tasks. Future work will extend validation to real-field datasets with natural variability.

We show that Out-of-Time-Ordered Correlator (OTOC) growth in the Quantum Kicked Rotor (QKR) quantifies information scrambling rather than entropy production. Numerical simulations reproduce the quadratic OTOC scaling at resonance (ℏeff=4π) and its suppression under detuning. Bitstreams derived from the evolving wavefunction reveal a nonmonotonic relationship between chaos and entropy: the resonant (maximally chaotic) regime exhibits lower measured entropy due to coherent phase correlations, whereas slight detuning enhances statistical uniformity. While Out-of-Time-Ordered Correlators quantify information scrambling rather than entropy production, we show that the strength of scrambling strongly constrains the amount of classical entropy that can be extracted after measurement.

This study addresses the need for effective extractive text summarization methods specifically designed for Indonesian news texts, driven by the increasing volume of digital information and persistent challenges related to semantic drift and coherence instability. A key challenge in extractive summarization lies in maintaining semantic fidelity while preserving informative content and ensuring computational efficiency, particularly for morphologically complex languages such as Indonesian. To address this issue, this research proposes a hybrid extractive summarization approach that integrates statistical methods, namely Term Frequency–Inverse Document Frequency (TF-IDF) and Latent Semantic Analysis (LSA), with Indonesian-specific transformer models, including IndoBERT and GPT, to enhance semantic similarity measurement during sentence selection. The proposed framework employs a weighted hybrid formulation (α·TF-IDF + β·LSA + γ·IndoBERT + δ·GPT), combined with cosine similarity computation and diversity-aware sentence selection. The approach is evaluated under a consistent experimental setting using the Indonesian subset of the XL-Sum dataset, with experiments conducted on TPU V2-8 infrastructure. Experimental results demonstrate competitive performance across the evaluated baseline components. The TF-IDF + LSA model achieved the highest ROUGE-1 score of 63.76%, while the TF-IDF + Cosine Similarity model showed balanced performance across ROUGE metrics with a cosine similarity score of 0.8538. IndoBERT + Cosine Similarity and GPT + Cosine Similarity achieved higher semantic similarity scores, with GPT reaching 0.8677. Overall, the proposed approach demonstrates substantial improvements over Indonesian extractive summarization baselines under the same dataset and evaluation protocol, highlighting the effectiveness of hybrid statistical-transformer integration for low-resource language summarization.

This study explores the application of Nadaraya-Watson Kernel Regression, a nonparametric statistical estimator within a machine learning (ML) framework, to model the thermal stability of Metal-Organic Frameworks (MOF) based on their structural descriptors. The method addresses the nonlinear and complex relationships often encountered in QSPR (Quantitative Structure-Property Relationship) studies, where conventional parametric models may fall short. Model performance was evaluated through 10-fold cross-validation and independent testing using R², RMSE, and MAE metrics. The best testing performance was achieved at a bandwidth of 0.1, yielding an R² of 0.99997, RMSE of 0.00039, and MAE of 0.000099. Visual analysis shows that smaller bandwidths enable the model to capture localized variations effectively, while larger bandwidths introduce over-smoothing and loss of critical structural information. Further feature analysis using Partial Dependence Plots (PDP) revealed that nitrogen atom count (nN) and heteroatom content (Het) positively influence thermal stability, while ligand count (Lig) has a negative effect. Zinc atom count (nZn) exhibits a nonlinear relationship, with a peak influence at moderate levels. These findings underscore the significance of kernel bandwidth tuning and feature interpretability in enhancing the predictive power of nonparametric models in material informatics.

Quantum Machine Learning (QML) has emerged as a promising interdisciplinary paradigm that integrates quantum computing and machine learning to address limitations of classical approaches in processing complex high-dimensional heterogeneous and imbalanced medical data. This study presents a systematic literature review of QML applications in healthcare based on 94 primary studies published between 2020 and April 2025. The review follows Kitchenham’s guidelines and the PRISMA framework to ensure rigor transparency and reproducibility. Seven research questions were formulated to examine publication trends sources datasets topics quantum model categories algorithms methods and evaluation metrics. Findings reveal a sharp increase in publication activity after 2023 indicating growing academic interest in healthcare-oriented QML. Medical imaging datasets were most frequently used followed by signal-based and tabular datasets with 86.5% of studies relying on public data. Hybrid classical-quantum approaches including CNN-QSVM Quantum Neural Networks and Variational Quantum Classifiers were dominant because they combine classical feature extraction with quantum-enhanced learning. However most studies were conducted in simulation environments rather than real hardware reflecting NISQ limitations including noise decoherence limited qubits and circuit depth constraints. Many results used small curated or imbalanced datasets raising concerns regarding generalizability reproducibility and clinical validity. Although numerous studies reported high accuracy standardized evaluation protocols external validation and uncertainty assessment remain limited. Overall QML in healthcare remains exploratory with substantial potential but significant barriers. This review highlights progress gaps and future directions emphasizing clinically representative datasets robust benchmarking interpretable noise-resilient models and real-hardware implementation to support trustworthy deployment in healthcare systems worldwide settings today.

Metal–organic frameworks (MOFs) are tunable porous materials that have been widely studied for gas adsorption, separation, and catalysis. Predicting and classifying the pore-limiting diameter (PLD) are valuable for rational MOF screening and design. Here, we develop an ensemble machine-learning pipeline: Random Forest (RF), XGBoost (XGB), and Light Gradient Boosting Machine (LGBM) to classify pore classes and regress PLD using a curated dataset derived from the Cambridge Structural Database. To address class imbalance across pore classes, we evaluate TomekLinks, Synthetic Minority Oversampling Technique (SMOTE), and SMOTETomek. Among the resampling strategies, RF provided the strongest overall performance, and the RF+TomekLinks configuration delivered the best balance between classification and regression. On the held-out test set, the best model achieved R^2 up to 0.999 with RMSE as low as 0.124 Å for PLD regression. Feature-importance analysis indicated that both metal-center elemental properties and linker-related topological descriptors contribute substantially to PLD prediction. Applicability-domain assessment using leverage/Williams-plot analysis shows that most test structures lie within the modeled domain, whereas out-of-domain cases are explicitly flagged. Overall, the proposed workflow enables accurate, domain-aware PLD estimation to support high-throughput MOF screening; nevertheless, external validation of independent MOF collections and uncertainty quantification remain important for broader deployment.

This study presents an enhanced Quantum Support Vector Regression approach with quantum kernels and virtual sampling for accurately predicting the formation energy of ABX3 perovskite materials. The model integrates quantum circuit expressibility and entanglement to improve prediction stability and representation capacity. Layer optimisation shows that a three-layer configuration achieves the best performance, balancing accuracy and computational cost. To address the challenges of limited data, virtual samples generated through kernel density estimation significantly enrich the dataset, increasing prediction accuracy to a determination coefficient of 0.99 with low errors, including root mean squared error of 0.087, mean absolute error of 0.053, and mean absolute deviation of 0.027. Comparative results show that the proposed quantum regression model outperforms classical approaches in terms of consistency and stability. At the same time, feature importance analysis based on Shapley values reveals that electronic properties and effective mass influence the formation energy. This hybrid quantum-classical machine learning framework demonstrates strong potential to accelerate the discovery of stable perovskite materials, reduce computational resources, and bridge the gap between classical and quantum paradigms in materials informatics. The proposed approach enables faster and more cost-efficient material innovation for energy and sustainability applications.

Class imbalance is a critical issue in binary classification of tabular data, where classical oversampling methods such as SMOTE often struggle in high-dimensional and non-linear settings. This study introduces QuantumEnhanced SMOTE (QES), a hybrid data augmentation framework that combines Variational Quantum Circuits (VQC) and Quantum Kernel Methods (QKM) to generate higher-quality synthetic minority samples. VQC facilitates non-linear interpolation in a quantum feature space, while QKM improves class separability using fidelitybased similarity measures. QES is implemented on quantum simulation platforms and evaluated on several imbalanced benchmark datasets, including one for perovskite ABO3 materials. Experimental results show that QES consistently outperforms SMOTE, SMOTE-Tomek, and SMOTE-ENN, particularly in AUC-PR, recall, and F1-score. These findings demonstrate that QES provides an effective, interpretable quantum-enhanced solution for imbalanced learning problems.

Investigating how a surface affects first N–H and N–N bond cleaving in hydrazine (N2H4) is a critical step towards designing a selective catalyst. In this study, the first N–H and N–N bond cleaving in hydrazine (N2H4) and hydroxyl (OH) coadsorption system on the Ni(211) surface is investigated using Density Functional Theory (DFT) based calculations. Effects of the vicinal steps and edges in Ni(211) on the H-bond formation and the selectivity of each bond-cleaving reaction are demonstrated. The product of N–H bond cleaving is shown to be less stable than N–N bond cleaving. Nevertheless, the activation energy of N–H bond cleaving is demonstrated to be significantly lower than that of N–N. In an ideal extended periodic system used in this study at specific coverage, the results demonstrate the effect of utilizing a corrugated surface in promoting first N–H bond cleaving instead of N–N.

Quantum computing is a computational process that utilizes quantum mechanics features, namely superposition, interference, and entanglement, in information processing, allowing computation to run in parallel. The advantage of quantum computing is that it solves complex problems whereas classical computing is impossible because it requires expensive computing costs. Object detection and recognition is a task of computer vision, where research in this field aims to improve the ability of computer algorithms to produce interpretations of visual information. Humans easily analyze and describe the visual information received. However, unlike computer systems, they must learn and explore using machine learning from the visual information received to provide correct interpretations of visual information. This paper presents a systematic review of papers published from 2012 to 2024 to answer how far quantum object detection and recognition research has been conducted. The methodology of this review follows a systematic literature review such as the method proposed by Kitchenham et al. The selected primary studies amounted to 29 papers from four source digital libraries. The application of quantum algorithms is more often used to improve the performance of classical computing. The quantum model category consists of 3 types, namely pure quantum, hybrid classical-quantum, and quantum-inspired ML. Hybrid classical-quantum is the most discussed model and Quantum Convolutional Neural Network is the most frequently discussed algorithm or model in image classification from 2012 to 2024. Quantum algorithms show good results and can improve the performance of classical algorithms, although currently, the ability of quantum computing is not fully optimal because the development of quantum computers is still in the noisy intermediate-scale quantum era. However, with the current limited quantum computing capabilities, it can already outperform the capabilities of classical computing. Based on this, studies on quantum object detection and recognition need to be carried out so that when the full potential of quantum computing can be utilized, the user's capacity is competent.

This study investigated the corrosion inhibition potential of ionic liquid compounds using a QSPR-based machine learning predictive model combined with DFT calculations. The Gradient Boosting (GB) model was identified as the most effective predictor, demonstrating excellent accuracy with a high R² value of 0.98. Additionally, the model exhibited low RMSE (0.95), MAE (0.84), and MAD (0.94) values. The predicted corrosion inhibition efficiencies (CIE) for three new ionic liquid compounds (IL1, IL2, and IL3) were 88.95, 90.82, and 93.16, respectively, which aligned well with experimental data. By integrating DFT simulations into the data updating process, facilitated by machine learning, the approach proved invaluable for identifying new corrosion inhibitors. This work highlighted the continuous refinement of data related to the corrosion inhibition effects of ionic liquid compounds.

Tomato is one of the widely available horticultural products and holds significant economic value in Indonesia. However, its productivity is often disrupted by various leaf diseases. This study aims to compare the performance of three CNN architectures—DenseNet121, Xception, and MobileNetV2—in classifying tomato leaf diseases. The dataset used consists of 10,000 balanced images across ten classes: Bacterial Spot, Septoria Leaf Spot, Early Blight, Late Blight, Mosaic Virus, Yellow Leaf Curl Virus, Leaf Mold, Target Spot, Spider Mites Two-Spotted Spider Mite, and Healthy. All images were resized to 224x224 pixels and divided into 80% training data and 20% test data. Augmentation techniques were applied to balance the data across classes. Experimental results show that the Xception architecture outperforms the other models, achieving an accuracy of 98.79%, with a precision of 98.80%, recall of 98.79%, and an F1-Score of 98.78%. These findings indicate that the Xception model is highly effective for plant disease classification and is suitable for implementation in environments with limited resources.

This study proposes a quantum key distribution (QKD) method based on the BB84 protocol, enhanced with quantum entanglement to improve key integrity and eavesdropping detection. The protocol employs two complementary metrics: Quantum Bit Error Rate (QBER) and Entanglement Check Ratio (ECR). QBER measures errors in key bits when both parties use the same basis, serving as the primary intrusion detector, while ECR reflects the proportion of entangled qubit pairs maintaining correlation across cross-basis measurements, acting as an indicator of entanglement integrity. The Clauser-Horne-Shimony-Holt (CHSH) test further validates entanglement by confirming Bell inequality violations. Simulation results show that the protocol effectively detects eavesdropping and man-in-the-middle (MiTM) attacks, with QBER rising under attack and ECR/CHSH values reflecting statistical entanglement degradation. Evaluation under practical noise conditions, including photon loss and detector inefficiency, demonstrates that ECR reliably indicates entanglement integrity even when the number of sifted key bits is reduced, while QBER remains the primary indicator of intrusion. This dual-metric approach, reinforced by CHSH analysis, provides a robust framework for detecting external interference and validating quantum behavior internally. The proposed entangled BB84 protocol offers a secure and efficient QKD solution, with potential for extension to long-distance and more physically robust implementations.

Small datasets often lead to poor performance of data-driven prediction models due to uneven data distribution and large data spacing. One popular approach to address this issue is to use virtual samples during machine learning (ML) model training. This study proposes a Hamiltonian Circuit Virtual Sample Generation (HCVSG) method to distribute virtual samples generated using interpolation techniques while integrating the K-Nearest Neighbors (KNN) algorithm in model development. The Hamiltonian circuit is chosen because it doesn't depend on the distribution assumption and provides multiple circuits that allow adaptive sample distribution, allowing the selection of circuits that produce minimum errors. This method supports improving feature-target correlation, reducing the risk of overfitting, and stabilizing error values as model complexity increases. Applying this method to three datasets in material research (MLCC, PSH, and EFD) shows that HCVSG significantly improves prediction accuracy compared to conventional KNN and eight MTD-based methods. The distribution of virtual samples along the Hamiltonian circuit helps fill the information gap and makes the data distribution more even, ultimately improving the predictive model's performance.

The security of chaos-based cryptography depends on the quality of the chaotic sequences used as keystreams. However, conventional chaotic systems often suffer from limitations in entropy, sensitivity to initial conditions, and the quality of randomness. This study proposes an enhancement method based on Lyapunov Exponent (LE)-guided quantum XYZ rotational encoding applied to Lorenz, Chen, and Rossler systems. The chaotic sequences are further amplified using a single-qubit quantum circuit with Hadamard, RX, RY, and RZ rotations. The evaluation conducted before (classic) and after the transformation (quantum) includes key aspects: LE, Shannon entropy, autocorrelation, bifurcation diagram, trajectories, and NIST SP 800-22 statistical testing. To ensure robust statistical assessment, the chaotic sequences were generated using a Multiple-Run Strategy with varying initial conditions to construct bitstreams suitable for NIST analysis. The results demonstrate significant improvements: LE increased up to tenfold, entropy values approached theoretical limits, autocorrelation decreased notably, and all sequences successfully passed the NIST tests under recommended parameters. These findings affirm the potential of LE-guided quantum rotations as an effective strategy for enhancing chaotic dynamics and improving the statistical security of cryptography keystreams.

Data security has become a growing priority due to the increasing frequency of cyber-attacks, necessitating the development of more advanced encryption algorithms. This paper introduces Single Qubit Quantum Logistic-Sine XYZ-Rotation Maps (SQQLSR), a quantum-based chaos map designed to generate one-dimensional chaotic sequences with an ultra-wide parameter range. The proposed model leverages quantum superposition using Hadamard gates and quantum rotations along the X, Y, and Z axes to enhance randomness. Extensive numerical experiments validate the effectiveness of SQQLSR. The proposed method achieves a maximum Lyapunov exponent (LE) of approximate to 55.265, surpassing traditional chaotic maps in unpredictability. The bifurcation analysis confirms a uniform chaotic distribution, eliminating periodic windows and ensuring higher randomness. The system also generates an expanded key space exceeding 1040, enhancing security against brute-force attacks. Additionally, SQQLSR is applied to image encryption using a simple three-layer encryption scheme combining permutation and substitution techniques. This approach is intentionally designed to highlight the impact of SQQLSR-generated chaotic sequences rather than relying on a complex encryption algorithm. The encryption method achieves an average entropy of 7.9994, NPCR above 99.6%, and UACI within 32.8%-33.8%, confirming its strong randomness and sensitivity to minor modifications. The robustness tests against noise, cropping, and JPEG compression demonstrate its resistance to statistical and differential attacks. Additionally, the decryption process ensures perfect image reconstruction with an infinite PSNR value, proving the algorithm's reliability. These results highlight SQQLSR's potential as a lightweight yet highly secure encryption mechanism suitable for quantum cryptography and secure communications.

Multifactorial colorectal cancer is a frequently diagnosed disease that is a leading cause of cancer-related mortality worldwide. Optimizing the patient's chances of recovery requires early detection. This research aims to implement a Convolutional Neural Network (CNN) model that can classify colon endoscopic images into three categories: normal, polyp, and cancer, using the VGG-16 architecture. This study employed the following techniques: K-means clustering to manage outliers, data augmentation to enhance the diversity of datasets, and Pearson correlation analysis to confirm the relationship between the augmentation and initial datasets. Data processing steps resulted in an enhanced dataset. Compared to models trained on the initial dataset, the results demonstrated a significant increase in accuracy and generalizability for CNN models trained on the enhanced dataset, which included outlier handling, augmentation, validation, and class balancing. The model's performance evaluation showed an accuracy of 86 %, with notable improvements in F1-score and recall for the cancer class. These findings indicate that the model better-classified images after dataset enhancement.