To construct a prediction model for fracture reduction fixator therapy using the multi-modal multi-label classification (MMC) method. Medical record data of 818 orthopedic patients from 2019 to 2023 were collected. Medical image features were extracted using the VGG19 network, text features of TCM four diagnostic methods (Inspection, Auscultation & Olfaction, Inquiry, Palpation) were extracted via the MiniLM model, and clinical case features were extracted through a fully connected neural network. After fusing the multi-modal information, multi-label therapy prediction was achieved using a linear layer. Experimental results on the clinical multi-modal dataset showed that the MMC method performed excellently in terms of subset accuracy(SA), accuracy(Acc), precision, and F1-score, reaching 0.661, 0.856, 0.897, and 0.899 respectively. When the image modality and text modality were removed, the model performance decreased by an average of 8.1% and 2.4% respectively, while the hamming loss(HL) increased by 21.1% and 5.6% respectively. The fracture reduction fixator therapy prediction model constructed in this study can effectively fuse multi-modal data, accurately predict personalized treatment plans for patients, and significantly improve the accuracy and reliability of treatment decisions. It provides a new solution for the digitalization and intellectualization of Traditional Chinese Medicine(TCM) in fracture treatment and has important clinical application prospects.

Community detection provides a principled lens on mesoscale organization in functional brain networks, yet many widely used methods presume assortative structure and depend on arbitrary thresholding, which complicates the selection of the community countK. We conducted a systematic benchmark of three assumption lean approaches that operate directly on weighted functional connectivity matrices: the Weighted Stochastic Block Model, Spectral Clustering, and K-means. Performance was assessed on synthetic networks with known ground truth and on three neuroimaging cohorts spanning development, namely the Human Connectome Project, Washington University 120, and the Baby Connectome Project. We compared strategies for choosingK, including post hoc indices such as silhouette, Calinski–Harabasz, C index, modularity, variation of information, Normalized Mutual Information, and zRand, together with a likelihood-based criterion for the Weighted Stochastic Block Model that uses bootstrap confidence intervals for differences in log likelihood between successive values ofK. In simulations all methods recovered stable partitions, but the post hoc indices favored incorrect values ofKunder weak signal and nonassortative mixing. In adult datasets the indices do not yield a unique optimum, whereas the likelihood-based criterion selects a parsimonious range centered nearK= 11, which is consistent with established sensory and association systems. In infants and toddlers, the same procedure supports a largerKaround 15 and reveals developmentally distinct mesoscale architecture, including anterior and posterior subdivisions within default mode and fronto parietal systems. A consensus relabeling scheme based on Hungarian matching with Hamming distance further stabilizes solutions across runs and across values ofK. Overall, threshold free weighted methods mitigate assortative bias and the likelihood-based comparison provides a reproducible path to selectingK.

Personalized survival benefit estimation from living donor liver transplantation with a novel machine learning method for confounding adjustment.

The rapid evolution of artificial intelligence-generated content (AIGC) has accelerated educational digitalization, yet its misuse—such as automated paper generation—poses unprecedented threats to academic ethics. This phenomenon not only undermines students’ critical thinking but also disrupts the fairness of educational evaluation. However, its improper application in the university context, such as students using AIGC tools to batch-generate “AI ghostwritten papers” and assignments, is increasingly eroding the foundation of academic integrity and weakening students’ critical thinking and originality. There is an urgent need for effective digital governance tools to safeguard the health and sustainability of the educational ecosystem. This study focuses on the core demands of digital technology empowering the sustainable development of education and proposes a multi-modal similarity model framework for AIGC content detection in educational scenarios. On the one hand, this model builds a three-dimensional multi-modal similarity recognition system, innovatively integrating text, formula, and chart information. On the other hand, combined with Hamming distance calculation, it achieves the deep integration of multi-source information and the identification of abnormal content through deep semantic feature extraction, structured content parsing, binary tree index structure, and local invariant feature matching technologies. Experimental verification has demonstrated the model’s accuracy and scalability. It provides digital technology support and practical paths for empowering educators and ethical governance in education to maintain a healthy academic ecosystem and educational equity, empowering learners to return to deep thinking to protect and promote the sustainable development of their core competencies such as critical thinking and innovation, and building a trustworthy, future-oriented sustainable education system.

Developing a multi-label learning model to predict major adverse cardiovascular events in patients with unstable angina pectoris: A prospective cohort study.

Inspired by the human retina, retinomorphic systems achieve efficient near-sensor processing by tightly integrating sensing, memory, and computing. However, unlike biological vision, which evolved without selective pressure for data confidentiality, artificial edge systems face critical security demands. Therefore, next-generation hardware must extend beyond biological mimicry by combining bio-inspired efficiency with cryptographic capabilities. Here, a compact, multifunctional wafer-scale monolithic 3D (M3D) architecture is proposed for secure in-memory processing of optically acquired visual data. Integrating quantum dot-sensitized phototransistors with stacked high-density vertical resistive random-accessmemories (VRRAMs) provides multi-domain entropy sources, generating physical unclonable function (PUF) keys with ≈50% inter-device variability. Multi-layer encryption using functionally independent PUF keys enhances cryptographic resilience through key diversity. Concurrently, M3D ternary content-addressable memory (TCAM) array, implemented with wide-bandgap IGZO transistors, achieves high sensing margin (≈1.58 × 105), along with 9.61× area efficiency and 6.25× energy-delay product improvements over planar designs. Notably, M3D sensory and TCAM systems support near-sensor hashing and in-memory Hamming distance computation directly on encrypted data, enabling application-specific homomorphism with a 94.1% similarity preservation rate. Comparable classification accuracy for plaintext and encrypted hash inputs further underscores the potential of M3D-integrated platforms for secure, privacy-preserving machine vision at the edge.

Accurate extraction of Population, Intervention, Comparison, and Outcome (PICO) elements from clinical texts is essential for supporting evidence-based medicine, particularly in cardiology where clinical data complexity presents significant challenges. This study investigates the comparative effectiveness of three contextual embedding models—BioBERT, PubMedBERT, and SciBERT—integrated with a Bidirectional Long Short-Term Memory (BiLSTM) architecture for multi-label PICO element classification on heart disease datasets. The experimental framework involved pre-processing clinical sentences, transforming them into contextual embeddings, and classifying PICO elements using BiLSTM-based sequence modeling. Evaluation was conducted using five key metrics: accuracy, precision, recall, F1-score, and hamming loss, supplemented by confusion matrix analysis for each PICO element. Results demonstrate that the BioBERT-BiLSTM model achieved superior performance, with an accuracy of 73.89%, F1-score of 78.54%, precision of 81.60%, and recall of 76.64%. PubMedBERT-BiLSTM exhibited the highest precision (84.12%) but lower recall, while SciBERT-BiLSTM produced slightly inferior results overall. These findings confirm the importance of using domain-specific embeddings, particularly models pre-trained on biomedical corpora, to improve classification accuracy in specialized clinical text tasks. This study concludes that the BioBERT-BiLSTM combination offers a reliable approach for automated PICO element extraction in the cardiology domain, contributing to the development of more accurate and efficient clinical decision-support systems

Let p be a prime and Fp a finite field of order p. This paper investigates cyclic codes over the ring Rp2,u=Zp2+uZp2 of order p4, where the nilpotent element u satisfies u2=0 and pu≠0. The condition u2=0 with pu≠0 is crucial, as it creates a nontrivial interaction between the components of the ring, allowing the construction of new codes with enhanced structural and distance properties. We provide explicit generating sets for cyclic codes over Rp2,u and study fundamental parameters such as their rank and Hamming distance. In the case gcd(n,p)=1, we show that cyclic codes can be generated by just two polynomials, which allows a complete determination of their rank and minimal Hamming distance distributions. Furthermore, using the Gray map from Rp2,u to Fp4, we construct all but one of the ternary optimal codes of length 12 as images of cyclic codes over R32,u, with computations verified using the Magma system.

Sparse numerical datasets are dominant in fields such as applied mathematics, astronomy, finance, and healthcare, presenting challenges due to their high dimensionality and sparse distribution. The predominance of zero values complicates optimal feature selection, making data analysis and model performance more complex. To overcome this challenge, this study introduces a deep learning-based algorithm, Hybrid Stacked Sparse Autoencoder (HSSAE), which integrates {text{L}}_{1} and {text{L}}_{2} regularization with binary cross-entropy loss to improve feature selection efficiency, where {text{L}}_{1} regularization penalizes large weights, simplifying data representations, while {text{L}}_{2} regularization prevents overfitting by limiting the total weight size. Additionally, the dropout technique enhances the algorithm’s performance by randomly deactivating neurons during training, avoiding over-reliance on specific features. Meanwhile, batch normalization stabilizes weight distributions, reducing computational complexity and accelerating the convergence. The proposed algorithm, HSSAE, was evaluated against traditional classifiers, including Decision Tree, Random Forest, K-Nearest Neighbors, and Naïve Bayes, as well as deep learning-based models, such as Convolutional Neural Network, Long Short-Term Memory, and Stacked Sparse Autoencoder, in terms of Precision, Recall, Accuracy, F1-score, AUC, and Hamming Loss. Quantitatively, the proposed algorithm, HSSAE, was tested on two different sparse datasets, demonstrating superior performance with the highest accuracy of 89% on the health indicator dataset and 93% on the EHRs diabetes prediction dataset, respectively, and outperforming competing classifiers. The proposed algorithm, HSSAE, extracts features effectively and enhances robustness, making it well-suited for sparse data applications, particularly in healthcare, where high prediction accuracy is crucial.

With the development of remote sensing technologies, the volume of remote sensing data is growing dramatically, making efficient management and retrieval of large-scale remote sensing images increasingly important. Recently, deep hashing for content-based remote sensing image retrieval (CBRSIR) has attracted significant attention due to its computational efficiency and high retrieval accuracy. Although great advancements have been achieved, the imbalance between easy and difficult image pairs during training often limits the model’s ability to capture complex similarities and degrades retrieval performance. Additionally, distinguishing images with the same Hamming distance but different categories remains a challenge during the retrieval phase. In this paper, we propose a novel deep relevance hashing (DRH) for remote sensing image retrieval, which consists of a global hash learning model (GHLM) and a local hash re-ranking model (LHRM). The goal of GHLM is to extract global features from RS images and generate compact hash codes for initial ranking. To achieve this, GHLM employs a deep convolutional neural network to extract discriminative representations. A weighted pairwise similarity loss is introduced to emphasize difficult image pairs and reduce the impact of easy ones during training. The LHRM predicts relevance scores for images that share the same Hamming distance with the query to reduce confusion in the retrieval stage. Specifically, we represent the retrieval list as a relevance matrix and employ a lightweight CNN model to learn the relevance scores of image pairs and refine the list. Experimental results on three benchmark datasets demonstrate that the proposed DRH method outperforms other deep hashing approaches, confirming its effectiveness in CBRSIR.

R1.C1: The dairy industry is a cornerstone of global food production and economic development; yet, its productivity is frequently hindered by common bovine health issues, including lameness, mastitis, metritis, and foot-and-mouth disease. These conditions not only affect milk yield but also pose significant challenges to maintaining animal welfare, highlighting the urgent need for intelligent, data-driven monitoring systems. R1.C2: In response to this critical need, this research proposes a machine learning (ML)-based framework for the early detection of such bovine events and diseases through multi-label classification. R1.C3: The system identifies estrus, calving, lameness, mastitis, and acidosis by analyzing key behavioral metrics derived from sensor-based monitoring, including feeding duration, resting periods, locomotion patterns, and aggregated activity data. R1.C4: In the context of multi-label bovine disease prediction, the combination of SMOTE and Classifier Chains is particularly crucial and synergistic due to the nature of the data and the interdependent relationships among the labels. R1.C5: The system was tested using a large dataset of 2.35 million records of livestock behavioral metrics. R1.C6: Among the six machine learning models investigated, the classifier chain configuration utilizing an Extra Tree Classifier consistently demonstrated superior performance, achieving a remarkable 97% subset accuracy, 96% recall, 95% precision, 96% F1-score, and a minimal Hamming loss of 0.04. Therefore, it is evident that classifier chains combined with oversampling techniques can capture label correlations and improve prediction performance compared to standard binary relevance approaches.

We develop algorithms for computing permutation polynomials (PPs) using normalization, so-called F-maps and G-maps, and the Hermite criterion. This allows for a more efficient computation of PPs for larger degrees and for larger finite fields. We use this to improve some lower bounds for , the maximum number of permutations on n symbols with a pairwise Hamming distance of D.

Heterogeneity in health care pathways preceding the classical recognition of adult-onset multiple sclerosis: A multichannel state sequence analysis.

Background and Objective: This study investigates the characterization of body odor signatures for early disease detection, aiming to demonstrate the feasibility of using simulated olfactory profiles within a computational diagnostic framework. The motivation arises from the growing interest in non-invasive diagnostic alternatives based on volatile organic compounds (VOCs) emitted by the human body. Materials and methods: A simulation-based approach was implemented using validated VOC datasets to construct binary odor profiles. These profiles were encoded as binary vectors, with each bit indicating the presence or absence of a specific compound. A simplified binary matching algorithm, excluding mutation and crossover operations, was employed to simulate pattern matching. The Hamming distance was used as the fitness function to quantify the similarity between profiles. Results and Discussion: The results indicate that the simplified binary matching algorithm reliably identified pathological odor profiles, producing high similarity scores with reference signatures. Despite the absence of conventional genetic operators, the method consistently converged to optimal or near-optimal matches. These findings emphasize the potential of binary odor encoding for distinguishing between healthy and pathological states, underscoring the robustness of the simplified computational framework. Conclusion: This work presents a novel and interpretable computational model for olfactory-based disease detection using simulated binary VOC patterns. It supports the development of low-cost, non-invasive diagnostic tools in medical contexts. Future research should explore extending the method by incorporating continuous VOC encoding, integrating evolutionary operators, and validating the results with semi-experimental or clinical data.

This paper expands the Mallows model for use in combinatorial domains. The Mallows model is a popular distribution used to sample permutations around a central tendency but requires a unique normalizing constant for each distance metric used in order to be computationally efficient. In this paper, closed-form expressions for the Mallows model normalizing constant are derived for the Hamming distance, symmetric difference, and the similarity coefficient in combinatorial domains. Additionally, closed-form expressions are derived for the normalizing constant of the weighted Mallows model in combinatorial domains. The weighted Mallows model increases the versatility of the Mallows model by allowing granular control over likelihoods of individual components in the domain. The derivation of the closed-form expression results in a reduction of the order of calculations required to calculate probabilities from exponential to constant.

Let G be a graph with n vertices and m edges. Let V(G) = {v1,v2,...,vn} be the vertex set of G. Thestring s(vi) is the row in the incidence matrix of G corresponding to the vertex vi, which is an m-tuple in Zm 2 . The Hamming matrix H(G) = [hij] of a graph G is an n×nmatrix, whose (i,j)-th entry is the Hamming distance between the strings s(vi) and s(vj). The Hamming energy HE(G) of a graph G is the sum of the absolute values of the eigenvalues of H(G). Recently the Hamming energy is introduced and obtained bounds for it in terms of the Hammingindexandobservedits predictive potentiality by correlating the physicochemical properties of molecules. In this paper we give the better bound for Hamming energy in terms of number of vertices and edges. Also obtain the largest eigenvalue of the Hamming matrix of a regular graph. Further obtain explicitly the eigenvalues of the Hamming matrix and Hamming energy of a complete bipartite graph

Legal violations in Indonesia, particularly those under the Criminal Code (KUHP) and the Information and Electronic Transactions Law (UU ITE), are often difficult for the general public to interpret due to the complexity of legal language and article structures. This research aims to build a multilabel classification model that can automatically identify relevant legal articles from user-provided case descriptions. Two models were developed and compared: Bidirectional Long Short-Term Memory (Bi-LSTM) and IndoBERT. Using a manually labeled dataset, both models were evaluated through accuracy, F1-score, and Hamming Loss metrics, as well as 5-fold cross-validation. The results showed that IndoBERT outperformed Bi-LSTM with an average accuracy of 97% and a Hamming Loss of 0.027. However, t-test analysis revealed no statistically significant difference in F1-scores, indicating that both models have comparable effectiveness in capturing multiple labels. A confusion matrix analysis further identified patterns of misclassification in semantically similar articles. This study demonstrates the potential of NLP and deep learning to support legal awareness and provide the public with easier access to legal information.

This paper proposes a Recurrent Non-linear Autoregressive Network with Exogenous inputs (RNARXNet) for Solar Power Forecasting (SPF) using time series data. Initially, input time series data is acquired from the Solar Power Generation dataset. Next, technical indicators such as Stochastic Oscillator (STOCH), Aroon (AR), Williams % R (WillR), Time Series Forecast (TSF), Moving Average Convergence Divergence (MACD), and Rainbow Moving Average are extracted. Afterwards, feature selection is done based on Hamming distance. Lastly, SPF is carried out utilizing RNARXNet which is newly modelled by integrating the Recurrent Radial Basis Function network (RRBFN) with a Non-linear Autoregressive Network with Exogenous inputs (NARX). The model obtained a normalized Mean Absolute Error (MAE) of 0.160, normalized Mean Square Error (MSE) of 0.352, normalized R 2 of 0.933, and normalized Root Mean Square Error (RMSE) of 0.594. These values show that RNARXNet is reliable and effective for precise SPF.

Machine learning strategies for multi-label pre-diagnosis of diseases with superficial data.

Consistent and Specific Hashing for image set classification.