Topic evolution and stance dynamics are deeply intertwined in online social media, shaping the fragmentation and polarization of public discourse.Yet existing dynamic topic models and stance analysis approaches usually consider these processes in isolation, relying on abstractions that lack interpretability and agent-level behavioral fidelity.We present stance and topic evolution reasoning framework (SPARK), the first LLM-based multi-agent simulation framework for jointly modeling the co-evolution of topics and stances through natural language interactions.In SPARK, each agent is instantiated as an LLM persona with unique demographic and psychological traits, equipped with memory and reflective reasoning.Agents engage in daily conversations, adapt their stances, and organically introduce emergent subtopics, enabling interpretable, fine-grained simulation of discourse dynamics at scale.Experiments across five real-world domains show that SPARK captures key empirical patterns-such as rapid topic innovation in technology, domain-specific stance polarization, and the influence of personality on stance shifts and topic emergence.Our framework quantitatively reveals the bidirectional mechanisms by which stance shifts and topic evolution reinforce each other, a phenomenon rarely addressed in prior work.SPARK provides actionable insights and a scalable tool for understanding and mitigating polarization in online discourse.Code and simulation resources will be released after acceptance.

The analysis of Cardiotocography (CTG) signals is often hindered by challenges such as limited data availability and label imbalance, which can undermine the performance of deep learning models. To address these issues, we present CTGDiff, a novel conditional diffusion model designed for generating synthetic Fetal Heart Rate (FHR) and Uterine Contraction (UC) signals. CTGDiff leverages both Phase-Rectified Signal Averaging (PRSA) spectrograms and UC as conditioning inputs for FHR, and integrates time encoding, condition generation from PRSA features, and residual blocks with dilated convolutions to capture both temporal dynamics and long-range dependencies. Extensive experiments, both qualitative and quantitative, demonstrate the model’s ability to synthesize high-quality CTG signals. In comparison with GANs and image-based diffusion models, CTGDiff achieves superior signal fidelity and distribution similarity for FHR, as indicated by metrics such as a 0.004 maximum mean deviation (MMD), 0.646 percent root mean square difference (PRD), 3.951 relative entropy (RE), and 0.291 Frechet distance (FD). Expert evaluations confirm that the model can generate both normal and abnormal CTG signals with high accuracy, conditioned on specific input data. These results underscore the potential of diffusion models for a wide range of applications in biomedical time series analysis, including signal synthesis, imputation, and noise reduction.

Accurate segmentation and landmark detection are essential for medical image analysis. However, existing methods often struggle with generalization and robustness, particularly under sparse annotation conditions. This study introduces SSP-MAEFNet, a novel self-supervised framework designed to address these limitations. The model integrates a lightweight backbone network, LR-ASPP-MobileNetV3, for efficient feature extraction, along with a Deep Separable Task-Specific (DSTS) module that mitigates feature entanglement between segmentation and landmark detection tasks. By leveraging a self-supervised random masking and reconstruction strategy, the framework effectively captures temporal and spatial patterns, enhancing its generalization across datasets with minimal supervision. Extensive experiments on the EchoNet-Dynamic dataset demonstrate that SSP-MAEFNet achieves state-of-the-art performance, with a Dice score of 93.54% and an OKS of 81.30%, outperforming existing models in both tasks. Its robust performance under sparse annotation highlights its potential for clinical integration, especially in resource-constrained environments.

Sleep staging is imperative for evaluating sleep quality and diagnosing sleep disorders. Extant sleep staging methods with fusing multiple data-views of physiological signals have achieved promising results. However, they remain neglectful of the relationship among different data-views at different feature scales with view position-alignment. To address this, we propose a novel cross-view alignment network, termed cVAN, utilising scale-aware attention for sleep stages classification. Specifically, cVAN principally incorporates two sub-networks of a residual-like network which learn spectral information from time-frequency images and a transformer-like network which learns corresponding temporal information. The prime advantage of cVAN is to adaptively align the learned feature scales among the different data-views of physiological signals with a scale-aware attention by reorganizing feature maps. Extensive experiments on three public sleep datasets demonstrate that cVAN can achieve a new state-of-the-art result, which is superior to existing counterparts.

The accurate diagnosis of multiple thoracic diseases through chest X-ray (CXR) images is a challenging yet crucial task in the medical field. Deep learning approaches have shown promise in assisting clinicians in this endeavor. In this paper, we introduce CvTGNet, a novel framework that leverages Convolutional Vision Transformer (CvT) and Graph Convolutional Network (GCN) to enhance CXR diagnosis. CvTGNet is designed to harness the super generalization capability of CvT, a hybrid architecture that combines Convolutional Neural Networks (CNNs) and Transformers. We employ GCN to explore the co-occurrence relationships among thoracic diseases, allowing the model to gain insights into intricate pathological connections that might be overlooked by traditional methods. This information guides the CvT model in making more accurate multi-label pathological classifications. We conducted extensive experiments on two prominent CXR datasets, ChestX-Ray14 and CheXpert, to evaluate the performance of CvTGNet. The results demonstrate that our proposed method consistently outperforms other state-of-the-art approaches in terms of Area Under the Receiver Operating Characteristic Curve (AUC-ROC) scores for multilabel pathological classification. We also conducted ablation experiments to understand the contribution of each component of CvTGNet. Overall, our CvTGNet framework presents a significant advancement in CXR diagnosis. By combining the strengths of CvT and CGN, we achieve remarkable accuracy in detecting multiple thoracic diseases from CXR images. This approach holds promise in enhancing medical diagnosis, enabling clinicians to make more informed decisions and improving patient outcomes.

Ensuring fetal health during pregnancy is critically dependent on precise Fetal Heart Rate (FHR) monitoring. A major challenge in this area is the limited availability of labeled FHR data, which poses a barrier to developing reliable automated analysis systems. To address this gap, our study introduces CTGGAN, a novel method employing Generative Adversarial Networks (GANs) to create synthetic, high-quality FHR signals. Specifically, CTGGAN integrates self-attention and residual modules within a Conditional GAN framework, fine-tuned to replicate the complex patterns characteristic of FHR data accurately. A notable feature of CTGGAN is its effective loss function, which combines Wasserstein distance with a gradient penalty to ensure training stability and enhance the authenticity of the generated signals. In performance metrics, Our method demonstrates the highest signal fidelity and distribution similarity, across five key measures: 0.215 maximum mean deviation (MMD), 0.012 sliced Wasserstein distance (SWD), 4.821 percent root mean square difference (PRD), 5.621 relative entropy (RE), and 0.614 Frechet distance (FD). This advancement in generating realistic FHR data with CTGGAN addresses critical issues like data insufficiency and class imbalance, thus advancing the field of prenatal healthcare technology. The code for CTGGAN is available at https://github.com/ijcnn2024/CTGGAN.

Cardiovascular diseases (CVDs) are the leading cause of global morbidity and mortality, necessitating the precise and continuous monitoring of blood pressure for proactive management. Our study presents the AMRUNet: a novel network designed exclusively for PPG-only, noninvasive, cuff-less blood pressure estimation. The network innovates upon the U-Net architecture, integrating a MultiRes Block for detailed multi-scale feature fusion and a residual block to mitigate the issue of vanishing gradients. An attention mechanism is further employed to selectively enhance salient features within the PPG signal. Our PPG-only AMRUNet demonstrates exceptional performance in translating PPG data into accurate ABP waveforms, achieving mean absolute errors (MAE) that comply with the standards of both the British Hypertension Society (BHS) and the Association for the Advancement of Medical Instrumentation (AAMI). Our method demonstrates highest MAE for both systolic blood pressure (SBP) and diastolic blood pressure (DBP), achieving a 2.85 MAE for SBP and 1.79 MAE for SBP among competing models. The model’s proficiency in precisely estimating systolic and diastolic blood pressure, along with its ability to reconstruct continuous ABP waveforms, contributes to reliable and trustworthy medical decision-making systems. The code for AMRUNet can be accessible at https://github.com/ijcnn2024/AMRUNet.

Stance detection, which aims to identify public opinion towards specific targets using social media data, is an important yet challenging task. With the proliferation of diverse multimodal social media content including text, and images multimodal stance detection (MSD) has become a crucial research area. However, existing MSD studies have focused on modeling stance within individual text-image pairs, overlooking the multi-party conversational contexts that naturally occur on social media. This limitation stems from a lack of datasets that authentically capture such conversational scenarios, hindering progress in conversational MSD. To address this, we introduce a new multimodal multi-turn conversational stance detection dataset (called MmMtCSD). To derive stances from this challenging dataset, we propose a novel multimodal large language model stance detection framework (MLLM-SD), that learns joint stance representations from textual and visual modalities. Experiments on MmMtCSD show state-of-the-art performance of our proposed MLLM-SD approach for multimodal stance detection. We believe that MmMtCSD will contribute to advancing real-world applications of stance detection research.

The advancement of machine vision systems necessitates efficient and accurate signal reconstruction methods to enhance real-time perception and decision-making capabilities. This paper introduces a Generalized Backtracking Regularization Adaptive Matching Pursuit (GBRAMP) algorithm, designed to reconstruct signals within machine vision systems using compressed sensing techniques. The GBRAMP algorithm improves upon existing methods by incorporating regularization for enhanced atom selection and a backtracking approach to accurately estimate sparsity, addressing the limitations of traditional convex optimization, greedy, and Bayesian reconstruction algorithms. The paper provides a comparative analysis of the GBRAMP algorithm against other prominent reconstruction techniques. Experimental results validate the GBRAMP algorithm's improved performance in terms of both reconstruction accuracy and computational speed, making it a competitive solution for the next generation of machine vision systems.

Financial time series is the main research direction of financial researchers due to its important economic value. However, owing to stochastic nature of financial markets, the data used by financial time series models need to have a series of features. This limits the use of financial forecasting models. A model for small datasets is proposed to address the above limitations. This model combines Adaptive Boosting(AdaBoost), K-Nearest Neighbors(KNN) and Long Short-Term Memory(LSTM) to accurately predict financial time series data by using a small number of features and data. Compared with classical machine learning models, the proposed model has the best prediction accuracy.

In recent years, machine learning has been widely used in data analysis of network engineering. The increasing types of model and data enhance the complexity of machine learning. In this paper, we propose a mathematical structure based on category theory as a combination of machine learning that combines multiple theories of data mining. We aim to study machine learning from the perspective of classification theory. Category theory utilizes mathematical language to connect the various structures of machine learning. We implement the representation of machine learning with category theory. In the experimental section, slice categories and functors are introduced in detail to model the data preprocessing. We use functors to preprocess the benchmark dataset and evaluate the accuracy of nine machine learning models. A key contribution is the representation of slice categories. This study provides a structural perspective of machine learning and a general method for the combination of category theory and machine learning.

Aspect-based sentiment analysis is an important task in fine-grained sentiment analysis, which aims to infer the sentiment towards a given aspect. Previous studies have shown notable success when sufficient labeled training data is available. However, annotating adequate data is labor-intensive, which sets substantial barriers for generalizing the sentiment predictor to the new domain. Two main challenges exist in cross-domain aspect-based sentiment analysis. One challenge is acquiring the domain-invariant knowledge; the other challenge is mining the syntactic-related words towards the aspect-term. In this article, we propose a transformer-based semantic-primary knowledge transferring network (TSPKT) for cross-domain aspect-term sentiment analysis, which utilizes semantic-primary knowledge as a bridge to enable knowledge transfer across domains. Specifically, we first build an S-Graph from external semantic lexicons, and extract the semantic-primary knowledge from the S-Graph. Second, AoaGraphormer is proposed to learn the syntactically relevant words towards the aspect-term. Third, we extend the standard biLSTM classifier to fully integrate the semantic-primary knowledge by adding a novel knowledge-aware memory unit (KAMU) to the biLSTM cell. Extensive experiments on six cross-domain setups demonstrate the superiority of TSPKT against the state-of-the-art baseline methods.

Aspect-term sentiment analysis (ATSA) is an important task that aims to infer the sentiment towards the given aspect-terms. It is often required in the industry that ATSA should be performed with interpretability, computational efficiency and high accuracy. However, such an ATSA method has not yet been developed. This study aims to develop an ATSA method that fulfills all these requirements. To achieve the goal, we propose a novel Sentiment Interpretable Logic Tensor Network (SILTN). SILTN is interpretable because it is a neurosymbolic formalism and a computational model that supports learning and reasoning about data with a differentiable first-order logic language (FOL). To realize SILTN with high inferring accuracy, we propose a novel learning strategy called the two-stage syntax knowledge distillation (TSynKD). Using widely used datasets, we experimentally demonstrate that the proposed TSynKD is effective for improving the accuracy of SILTN, and the SILTN has both high interpretability and computational efficiency.

With the development of technology and the expansion of social media, rumors spread widely and the rumor detection has gradually caused widespread concern. The early method of using handcrafted features has been eliminated due to inefficiency, and deep learning methods have been gradually adopted in recent years. However, most of the methods only consider content information such as text, which is often not enough for the specific field, rumor detection. Some studies take propagation rule into consideration, such as Kernel-based, RvNN. In addition, the structure formed via propagation of rumors and non-rumors have different properties. Compared with dynamic propagation, structure here is the final result of propagation and it's static and global. In order to enhance the structure information, we proposes a model that obtains textual, propagation and structure information. The model contains three components: Encoder, Decoder, and Detector. The encoder uses the efficient Graph Convolutional Network to regard the initial text as input and update the representation through propagation to learn text and propagation information. Then the encoded representation would be used for subsequent decoder which uses AutoEncoder to learn the overall structure information. Simultaneously, the detector utilizes the output of encoder to classify events as fake or not. These three modules are jointly trained to improve the model effect. We verified our method on three real-world datasets, and the results show that our method outperforms other state-of-the-art methods.

Classification problems in which the number of features (dimensions) is unduly higher than the number of samples (observations) is an essential research and application area in a variety of domains, especially in computational biology. It is also known as a high-dimensional small-sample-size (HDSSS) problem. Various dimensionality reduction methods have been developed, but they are not potent with the small-sample-sized high-dimensional datasets and suffer from overfitting and high-variance gradients. To overcome the pitfalls of sample size and dimensionality, this study employed variational autoencoder (VAE), which is a dynamic framework for unsupervised learning in recent years. The objective of this study is to investigate a reliable classification model for high-dimensional and small-sample-sized datasets with minimal error. Moreover, it evaluated the strength of different architectures of VAE on the HDSSS datasets. In the experiment, six genomic microarray datasets from Kent Ridge Biomedical Dataset Repository were selected, and several choices of dimensions (features) were applied for data preprocessing. Also, to evaluate the classification accuracy and to find a stable and suitable classifier, nine state-of-the-art classifiers that have been successful for classification tasks in high-dimensional data settings were selected. The experimental results demonstrate that the VAE can provide superior performance compared to traditional methods such as PCA, fastICA, FA, NMF, and LDA in terms of accuracy and AUROC.

Obstetric ultrasound examination of physiological parameters has been mainly used to estimate the fetal weight during pregnancy and baby weight before labour to monitor fetal growth and reduce prenatal morbidity and mortality. However, the problem is that ultrasound estimation of fetal weight is subject to populations’ difference, strict operating requirements for sonographers, and poor access to ultrasound in low-resource areas. Inaccurate estimations may lead to negative perinatal outcomes. We consider that machine learning can provide an accurate estimation for obstetricians alongside traditional clinical practices, as well as an efficient and effective support tool for pregnant women for self-monitoring. We present a robust methodology using a data set comprising 4,212 intrapartum recordings. The cubic spline function is used to fit the curves of several key characteristics that are extracted from ultrasound reports. A number of simple and powerful machine learning algorithms are trained, and their performance is evaluated with real test data. We also propose a novel evaluation performance index called the intersectionover-union (loU) for our study. The results are encouraging using an ensemble model consisting of Random Forest, XGBoost, and LightGBM algorithms. The experimental results show an loU of 0.64 between predicted range of fetal weight at any gestational age from the ensemble model and that from ultrasound. Comparing with the ultrasound method, the estimation accuracy is improved by 12%, and the mean relative error is reduced by 3%.

Fetal heart rate (FHR) is very significant to evaluate the status of fetus. However, based on traditional classification criteria is not accurate. With the rapid development of computer information technology, computer technology is vital for the analysis of FHR in electronic fetal monitoring (EFM). FHR is divided into three classes as: 1) normal; 2) suspicious; and 3) abnormal. Through the cooperation with the hospital, we got 4473 records, including 3012 normal, 1024 suspicious, 437 abnormal records by our EFM system. In order to improve the accuracy of fetal status assessment, high 1-D FHR records are divided into ten d-window segments, and then use convolutional neural network (CNN) to process the data in parallel. Finally, we use the voting method to determine the class of FHR records. We also made a comparative experiment, the feature extraction method based on basic statistics is used to extract the features of FHR. And then the features were applied as the input to support vector machine (SVM) and multilayer perceptron (MLP) to classify. According to the results of the experiment, the accuracy of classification of SVM, MLP, and CNN are 79.66%, 85.98%, and 93.24%, respectively.

Because of the discrepancy between different domains, the sentiment classifier trained in a source domain can't get a good performance in a target domain. Domain adaptation algorithms aim at solving such problems. One of the main algorithms aim at finding domain-invariable representations of inputs, which pays more attention to the common features of different domains and ignores the characteristics of samples themselves. In our paper, we propose a Fuzziness Based Domain-Adversarial Neural Network with Auto-Encoder (Fuzzy-DAAE). It not only uses a domain classifier to find domain-invariable features, but also uses an auto-encoder to reconstruct inputs to keep characteristics of samples. In order to introduce more supervised information of target samples, we also add unlabeled target samples and their predicted labels to the original training data according to their fuzziness and then retrain the whole model. Experiments on Amazon product reviews show that our proposed model has the best or comparative results compared with the existing models. It's worthwhile to notice that our model can be used in any other domain adaptation tasks, not limited to cross-domain sentiment classification.

Summary Recently, social network has been one of the biggest information exchange platforms of the Internet. Moreover, the users in social network used to watch videos through social network application. To provide a proper recommended video list, the video recommendation algorithm for social network is becoming a hot research issue. On one hand, more and more researchers introduce the concept of trust into video recommendation algorithms. However, most of them only select the trust friends based on the similarity and neglect the characteristics of social network. On the other hand, most previous video recommendation algorithms are only based on the number that a video is viewed to evaluate a video's quality. They do not make good use of the social relationship in social network and the video's reputation. This paper mainly focuses on the challenge that the effectiveness and performance of current video recommendation algorithm in social network cannot satisfy the users. In this paper, we propose a novel video recommendation algorithm based on the combination of video content and social network. Our proposed algorithm consists of the trust friends computing model and video's quality evaluation model. The trust friends computing method takes into account similarity between users, interaction between users, and the active degree of a user. In our video's quality evaluation model, we combine the acceptance ratio of a video with a video's reputation. The video can be given an appropriate rating score through this model. We design corresponding trust friends computing algorithm and video recommendation algorithm respectively for two proposed models. Our integral video recommendation algorithm consists of these two algorithms. The experimental results indicate that the performance and effectiveness of our algorithm are better than those of two classical video recommendation algorithms (i.e., user‐based collaborative filtering algorithm and TBR‐d algorithm), in terms of precision, recall and F1‐measure. Copyright © 2016 John Wiley & Sons, Ltd.

Using deep learning models to solve sentiment analysis of sentences is still a challenging task. Long short-term memory (LSTM) network solves the gradient disappeared problem existed in recurrent neural network (RNN), but LSTM structure is linear chain-structure that can’t capture text structure information. Afterwards, Tree-LSTM is proposed, which uses LSTM forget gate to skip sub-trees that have little effect on the results to get good performance. It illustrates that the chain-structured LSTM more strongly depends on text structure. However, Tree-LSTM can’t clearly figure out which sub-trees are important and which sub-trees have little effect. We propose a simple model which uses Rhetorical Structure Theory (RST) for text parsing. By building LSTM network on RST parse structure, we make full use of LSTM structural characteristics to automatically enhance the nucleus information and filter the satellite information of text. Furthermore, this approach can make the representations concerning the relations between segments of text, which can improve text semantic representations. Experiment results show that this method not only has higher classification accuracy, but also trains quickly.

We propose a Word-Topic Mixture(WTM) model to improve word representation and topic model simultaneously. Firstly, it introduces the initial external word embeddings into the Topical Word Embeddings(TWE) model based on Latent Dirichlet Allocation(LDA) model to learn word embeddings and topic vectors. Then the results learned from TWE are integrated in the LDA by defining the probability distribution of topic vectors-word embeddings according to the idea of latent feature model with LDA (LFLDA), meanwhile minimizing the KL divergence of the new topic-word distribution function and the original one. The experimental results prove that the WTM model performs better on word representation and topic detection compared with some state-of-the-art models.

Nonnegative matrix factorization is used extensively for feature extraction and clustering analysis. Recently many sparsity/sparseness constraints, such as L1penalty, are introduced for sparse nonnegative matrix factorization. Inspired by sparsity measures from linear regression model, this paper proposes to integrate nonnegative matrix factorization with another sparsity constraint, the elastic net. The experimental results of clustering analysis on three gene expression datasets demonstrate the effectiveness of the proposed method.