Yibo Wang , Xiuqin Ma , Hongwu Qin , Yuanyuan Chen , Jemal H. Abawajy
{"title":"Dynamic q-rung orthopair hesitant fuzzy decision-making model based on Banzhaf value of fuzzy measure","authors":"Yibo Wang , Xiuqin Ma , Hongwu Qin , Yuanyuan Chen , Jemal H. Abawajy","doi":"10.1016/j.asoc.2025.113036","DOIUrl":"10.1016/j.asoc.2025.113036","url":null,"abstract":"<div><div>Multi-attribute decision-making in dynamic fuzzy environments holds significant practical importance in fields. The changing fuzzy environmental factors and the hesitant psychology during the dynamic period commonly involve potential historical impacts of decisions and dynamic fuzzy intercorrelation. Quantitative indicators of attribute importance derived from fuzzy measures can account for intercorrelation but cannot ensure the dynamic feedback of historical decisions. For these reasons, the existing fuzzy decision-making methods have failure and poor stability when dealing with complex, hesitant dynamic fuzzy decision-making problems. Therefore, this paper aims to construct a dynamic decision-making model within the q-rung orthopair hesitant fuzzy(q-ROHF) environment. First, this paper derives the q-ROHF cross-entropy and four combinations of entropy formulas to derive the calculation method for the fuzzy measure of dynamic attributes. The Banzhaf value is used as a weighting metric to quantify the degree of impact of dynamic attributes to the decision result. Second, based on the Banzhaf value, a dynamic weighting algorithm incorporating a historical feedback mechanism and dynamic intercorrelation is proposed. Further, this paper derives a fuzzy preference relation(FPR) using the q-ROHF generalized fuzzy distance and presents an algorithm for updating the dynamic FPR matrix. Finally, a dynamic FPR-based AQM method is used to construct the dynamic decision-making model. The feasibility and effectiveness of the decision-making model in response to changing complex factors are demonstrated through two distinct real-world cases. Through robustness analysis and advantage analysis, it has been verified that the model exhibits superior dynamic decision-making stability and hesitant decision-making stability compared to existing models. The model can be a powerful tool for dealing with dynamic decision-making problems that involve hesitation and uncertainty.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113036"},"PeriodicalIF":7.2,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143746206","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Transcending modularity: A memetic algorithm combining triangle motif and edge information for community detection","authors":"Xiangyi Teng , Xinyue Luo , Jing Liu","doi":"10.1016/j.asoc.2025.113082","DOIUrl":"10.1016/j.asoc.2025.113082","url":null,"abstract":"<div><div>The research towards community detection plays a crucial role in revealing the topological structure and functional characteristics of complex networks. Nowadays, modularity and its variants are the most popular community quality evaluation metric applied in the detection of community structures. However, these modularity-based methods rarely consider higher-order structural information such as motifs in a network, which may lead to an incomplete and inaccurate understanding of the network. While some motif-based methods have been proposed, they suffer from resolution limitations and completely ignore lower-order structures. To bridge this gap and address community detection problem from a more hybrid view focusing on both lower-order and higher-order structures, this paper first proposes an adaptive hybrid-order modularity optimization function termed as TE-Modularity, which harmonizes triangle motif and edge information. It transcends traditional modularity by considering both lower-order and higher-order structures and can be applicable to all types of networks. In addition, we design a memetic algorithm called TE-MA, that uses TE-Modularity as the objective function to solve the community detection problem. A novel mutation operator based on triangle motifs is proposed, which can effectively accelerate the convergence of the proposed algorithm. Furthermore, we develop a new multipoint local search strategy, striking a good balance between the efficiency and quality of the algorithm. Through experiments conducted on different optimizers and comparisons with several state-of-the-art community detection methods, our approach's effectiveness and superiority are demonstrated on both real and synthetic networks.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113082"},"PeriodicalIF":7.2,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143759718","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yinghong Li , Yudong Yan , Zhuohao Tong , Yu Wang , Yinqi Yang , Mingze Bai , Dan Pu , Jiazheng Xie , Chuan Liu , Bo Li , Mingwei Liu , Kunxian Shu
{"title":"Efficient fine-tuning of small-parameter large language models for biomedical bilingual multi-task applications","authors":"Yinghong Li , Yudong Yan , Zhuohao Tong , Yu Wang , Yinqi Yang , Mingze Bai , Dan Pu , Jiazheng Xie , Chuan Liu , Bo Li , Mingwei Liu , Kunxian Shu","doi":"10.1016/j.asoc.2025.113084","DOIUrl":"10.1016/j.asoc.2025.113084","url":null,"abstract":"<div><div>The escalating computational costs of large language models (LLMs) have catalyzed the pursuit of more efficient alternatives, particularly in specialized domains like biomedicine. In this study, we propose BioQwen, a series of small-parameter biomedical bilingual (Chinese–English) multi-task models designed to mitigate the resource demands of LLMs while achieving high performance.</div><div>BioQwen is trained on carefully curated open-source biomedical datasets, employing a stringent preprocessing pipeline with thorough quality filtering and standardized formatting. Through an efficient two-stage fine-tuning strategy, BioQwen models with 0.5B, 1.5B, and 1.8B parameters attain competitive performance across a variety of comprehension and generative tasks. For comprehension tasks, BioQwen-1.8B achieves a Macro F1 score of 0.730 and a balanced accuracy of 0.802 on the BC5CDR dataset, surpassing the 7B-parameter Taiyi model’s scores of 0.685 and 0.757. In generative tasks, BioQwen delivers superior zero-shot results on the iCliniq dataset, outperforming all baselines across multiple metrics. Comparisons with established small-parameter LLMs (e.g., Llama3.2 1B) further substantiate the effectiveness of domain-specific fine-tuning.</div><div>Significantly, BioQwen’s reduced iteration time highlights its computational efficiency, and its quantized version demonstrates successful deployment on mobile devices, confirming its viability in resource-constrained settings. This study demonstrates the potential of strategically fine-tuned small-parameter LLMs to deliver resource-efficient, high-performing solutions for biomedical bilingual applications, expanding accessibility and usability in the field.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113084"},"PeriodicalIF":7.2,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yining Feng , Lu Wang , Jiarui Jin , Xianghai Wang
{"title":"A semi-supervised multi-source remote sensing image classification network based on adaptive pseudo-label generation","authors":"Yining Feng , Lu Wang , Jiarui Jin , Xianghai Wang","doi":"10.1016/j.asoc.2025.113055","DOIUrl":"10.1016/j.asoc.2025.113055","url":null,"abstract":"<div><div>Earth observation technology leveraging remote sensing (RS) imagery serves as a valuable non-contact detection method with broad applications in classification research. Hyperspectral (HS) image classification, while effective in various domains, faces challenges due to the unique characteristics of HS data. Fusion diverse RS data sources can mitigate redundancy and enhance classification efficiency. However, many deep learning approaches for multi-source RS classification rely heavily on abundant labeled data, which can be time-consuming and often impractical. To address the limitations in feature extraction and classification accuracy stemming from the scarcity of labeled multi-source RS image samples in complex scenes, we propose a novel semi-supervised multi-source RS image classification network based on adaptive pseudo-label generation (S2CNet-APG). This framework incorporates attention modules that effectively embed active RS features into HS features, enhancing performance through squeezing and excitation (SE) driven attention mechanisms. Our semi-supervised learning approach employs adaptive thresholds to manage the quantity of pseudo-labels derived from unlabeled samples, while maintaining the spatial consistency of the information to ensure quality. This dual strategy effectively balances the quantity and quality of pseudo-labels, enabling accurate classification with limited labeled samples and transitioning multi-source RS image classification from a supervised to a semi-supervised paradigm. We conducted extensive experiments on three real-world multi-source RS datasets, achieving superior results that validate the efficacy of the proposed method.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113055"},"PeriodicalIF":7.2,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738484","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Fanglin Liu , Qinghe Zheng , Xinyu Tian , Feng Shu , Weiwei Jiang , Miaohui Wang , Abdussalam Elhanashi , Sergio Saponara
{"title":"Rethinking the multi-scale feature hierarchy in object detection transformer (DETR)","authors":"Fanglin Liu , Qinghe Zheng , Xinyu Tian , Feng Shu , Weiwei Jiang , Miaohui Wang , Abdussalam Elhanashi , Sergio Saponara","doi":"10.1016/j.asoc.2025.113081","DOIUrl":"10.1016/j.asoc.2025.113081","url":null,"abstract":"<div><div>The Detection Transformer (DETR) has emerged as the dominant paradigm in the field of object detection due to its end-to-end architectural design. Researchers have explored various aspects of DETR, including its structure, pre-training strategies, attention mechanisms, and query embeddings, achiving significant progress. However, high computational costs limit the efficient use of multi-scale feature maps and hinder the full exploitation of complex multi-branch structures. We examine the negative impact of multi-scale features on the computational cost of DETRs and find that introducing long sequence data to the encoder is suboptimal. In this work, we aim to further push the boundaries of DETR’s performance and efficiency from the model structure perspective, thus developing the fusion detection Transformer (F-DETR) with heterogeneous scale multi-branch structure. To the best of our knowledge, this is the first explicit attempt to integrate multi-scale features into the end-to-end DETR structure. Specifically, we propose a multi-branch structure to simultaneously utilize feature maps at different levels, facilitating the interaction of local and global features. Additionally, we select certain joint latent variables from the interactive information flow to initialize the object container, a technique commonly used in query-based detectors. Experimental results show that F-DETR achieves a 43.9 % AP using 36 training epochs on the popular public COCO dataset. Furthermore, our approach demonstrates a better trade-off between accuracy and complexity compared to the original DETR.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113081"},"PeriodicalIF":7.2,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738105","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Wenbin Qian , Wenyong Ruan , Xiwen Lu , Wenji Yang , Jintao Huang
{"title":"Feature selection via label enhancement and neighborhood rough set for multi-label data with unbalanced distribution","authors":"Wenbin Qian , Wenyong Ruan , Xiwen Lu , Wenji Yang , Jintao Huang","doi":"10.1016/j.asoc.2025.113028","DOIUrl":"10.1016/j.asoc.2025.113028","url":null,"abstract":"<div><div>Multi-label learning has gained significant attention in classification tasks, but challenges remain in handling high-dimensional data. Although feature selection techniques can alleviate these issues, neglecting the unbalanced data distribution problem severely undermines the models’ accuracy. Furthermore, existing methods fail to account for the importance and correlation of labels. In this paper, we present a novel multi-label feature selection algorithm that addresses these issues through three innovations: (1) using <span><math><mi>k</mi></math></span>-nearest neighbors to capture local similarities in unbalanced data, (2) enhancing labels by converting them into distributions to enrich semantic information, and (3) introducing a new evaluation function to assess label correlations. A multi-criteria strategy is established to maximize feature-label relevance, minimize redundancy, and strengthen label correlations. Experimental results on fifteen multi-label datasets demonstrate the algorithm’s superiority over five state-of-the-art methods.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113028"},"PeriodicalIF":7.2,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143783102","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Yan Kang , Yue Peng , Dongsheng Zheng , Huadong Zhang , Xuekun Yang
{"title":"Multi-view framework for multi-label bioactive peptide classification based on multi-modal representation learning","authors":"Yan Kang , Yue Peng , Dongsheng Zheng , Huadong Zhang , Xuekun Yang","doi":"10.1016/j.asoc.2025.113007","DOIUrl":"10.1016/j.asoc.2025.113007","url":null,"abstract":"<div><div>The diversity and specific biological functions of bioactive peptides make them key regulators in various physiological processes and crucial contributors to the development of new anti-infective drugs. Although existing graph-based deep learning methods effectively model multi-label peptide representation, they often fail to incorporate multi-modal feature representation and extract multi-scale features from various views. To address these limitations, we present a multi-view framework for multi-label bioactive peptide classification based on multi-modal representation Learning by combining amino acid sequences and fusion molecular fingerprints. The peptide molecular graph is constructed by extracting the topological information and node features, respectively. Multi-view branches are designed by developing sequence-based and graph-based models to leverage their distinct and complementary strengths. Specifically, the protein language model ESM-2 is utilized to extract residue features from amino acid sequences deeply. Meanwhile, local features from molecular fingerprints are learned through a Filter Response Normalization layer and a Thresholded Linear Unit. At the same time, the Mamba module is innovatively employed to extract long-range dependencies and reduce time complexity. Our model demonstrates significantly enhanced and robust performance in multi-label bioactive peptide prediction tasks compared with state-of-the-art models, achieving 82.5% coverage, 80.9% precision and 80.3% accuracy on the MFBP dataset. Furthermore, visual analyses demonstrate that the model can effectively capture features from multiple views and highlight the interpretability of the model through the decision process.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113007"},"PeriodicalIF":7.2,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738289","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Mingjie Lu , Zhaowei Liu , Pengda Wang , Haiyang Wang , Dong Yang
{"title":"A Bayesian graph structure inference neural network based on adaptive connection sampling","authors":"Mingjie Lu , Zhaowei Liu , Pengda Wang , Haiyang Wang , Dong Yang","doi":"10.1016/j.asoc.2025.113018","DOIUrl":"10.1016/j.asoc.2025.113018","url":null,"abstract":"<div><div>Graph Neural Networks (GNNs) have drawn a lot of interest recently and excel in several areas, including node categorization, recommended systems, link prediction, etc. However, most GNNs by default observe graphs that accurately reflect the relationships between nodes. The feature aggregation of GNN is done by aggregating the neighbor nodes of the node. Therefore, observation graphs are not always compatible with the properties of GNNs. Unlike random regularization techniques that employ constant sampling rates or manually tune them as model hyperparameters. This study proposes a graph-structure learning network based on adaptive connection sampling. The core idea is to use the features generated by each layer of GNNs through adaptive sampling to generate a graph through the Bayesian method and realize the joint optimization of graph structure and adaptive connection sampling through iteration. This study conducts experiments on the data set to evaluate the effectiveness of this method. In the node classification task, the model improves performance by about 3.8% compared to the average of many baselines. It can be seen that learning graph structures is effective and inferring graphs is logical.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113018"},"PeriodicalIF":7.2,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143791935","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Fusing multi-granular-ball fuzzy information to detect outliers","authors":"Xinyu Su , Shitong Cheng , Dezhong Peng , Hongmei Chen , Zhong Yuan","doi":"10.1016/j.asoc.2025.113045","DOIUrl":"10.1016/j.asoc.2025.113045","url":null,"abstract":"<div><div>Outlier detection plays a critical role in data mining and machine learning, and its application value is widely recognized in several industries. However, despite the growing importance of outlier detection, many current outlier detection methods still rely on a single and fine-granularity processing paradigm. Not only does this paradigm lead to inefficient methods, but it also makes the methods vulnerable to noisy data. Furthermore, this processing paradigm ignores the potential multi-granularity information in the data, which may lead to an incomplete understanding of the intrinsic relations and patterns of the data. To further improve the performance of outlier detection, multi-granular-ball fuzzy information granules-based unsupervised outlier detection method (MGBOD) is proposed in this work. In our method, granular-balls with different granularity are first generated and the fuzzy binary relations between the granular-balls with respect to different attributes are computed. Subsequently, two attribute sequences are constructed based on the importance of the attributes. Then, multi-granular-ball fuzzy binary granular structures are constructed based on these two sequences. Finally, the outlier score of the granular-ball is defined by fusing these granules in the granular structures and mapped to the samples in the granular-ball. Experimental results show that, compared with recently proposed methods, our method demonstrates excellent outlier detection performance under a variety of public datasets. The code is publicly available at <span><span>https://github.com/Mxeron/MGBOD</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113045"},"PeriodicalIF":7.2,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143738478","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Evaluation of the anti-disturbance capability of fMRI-based spiking neural network based on speech recognition","authors":"Lei Guo , Chongming Li , Youxi Wu , Menghua Man","doi":"10.1016/j.asoc.2025.113069","DOIUrl":"10.1016/j.asoc.2025.113069","url":null,"abstract":"<div><div>The exterior electromagnetic noise can degrade the performance of neuromorphic hardware based on brain-inspired model. Therefore, enhancing the robustness of a brain-inspired model is a critical issue. However, the topology of a brain-inspired model lacks bio-plausibility. The purpose of this paper is to enhance the anti-disturbance capability of brain-inspired model under exterior electromagnetic noise by improving its bio-plausibility. In this paper, we propose a new spiking neural network (SNN) as a brain-inspired model called fMRI-SNN, in which the topology is constrained by functional magnetic resonance imaging (fMRI) data from the human brain, the nodes are Izhikevich neuron models, and the edges are synaptic plasticity models (SPMs) with time delay co-regulated by excitatory synapses and inhibitory synapses. Then, taken speech recognition (SR) as a case study, the recognition performance of fMRI-SNN is certified. To evaluate its anti-disturbance capability, the SR accuracy of fMRI-SNN under exterior electromagnetic noise is investigated, and is compared with SNNs with alternative topologies. To reveal its anti-disturbance mechanism, the neuroelectric characteristics, adaptive adjustment of synaptic plasticity, and dynamic topological characteristics of fMRI-SNN under exterior electromagnetic noise are discussed. The results indicate that the SR accuracy of fMRI-SNN under exterior electromagnetic noise is higher than that of SNNs with alternative topologies, and our discussion elucidates its anti-damage mechanism. Our results prompt that the brain-inspired model with bio-plausibility can enhance its robustness.</div></div>","PeriodicalId":50737,"journal":{"name":"Applied Soft Computing","volume":"175 ","pages":"Article 113069"},"PeriodicalIF":7.2,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143746374","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}