Information Sciences最新文献

Generative imputation of incomplete images: Leveraging multimodal information for missing pixel

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-04-02 DOI: 10.1016/j.ins.2025.122159

Qian Ma, Yuan Guo, Jinlei Zhang, Haochen Zhang, Shikai Guo, Bo Ning, Yu Gu, Yu Ge

{"title":"Generative imputation of incomplete images: Leveraging multimodal information for missing pixel","authors":"Qian Ma, Yuan Guo, Jinlei Zhang, Haochen Zhang, Shikai Guo, Bo Ning, Yu Gu, Yu Ge","doi":"10.1016/j.ins.2025.122159","DOIUrl":"10.1016/j.ins.2025.122159","url":null,"abstract":"<div><div>Missing pixels are a common issue in real-world images, arising from various factors such as hardware malfunctions, sensor errors, and other unforeseen circumstances. This prevalence of missing pixels has made incomplete image imputation a critical area of research, garnering attention both domestically and internationally. However, as the volume of data continues to grow, traditional imputation methods that rely exclusively on information from the target images are becoming less effective, particularly in scenarios where the proportion of missing pixels is high. To address this challenge, we propose a novel imputation model named MMIGAN (Multi-modal Imputation Generative Adversarial Network), which imputes incomplete images by leveraging not only the information from the images themselves but also additional information from corresponding texts. Specifically, MMIGAN is a GAN-based model where the generator G comprises a cross-modality feature learning subnet to extract multimodal features and an MV imputation subnet to output the imputed images. Meanwhile, the discriminator D attempts to distinguish between real (observed) and fake (imputed) pixels to enhance imputation accuracy. We conducted extensive experiments on the Flickr8k, Flickr30k, and COCO datasets, demonstrating that MMIGAN surpasses state-of-the-art methods in image inpainting tasks. Under varying missing rates, the peak performance improvements across these datasets reached 52.5%, 61.0%, and 54.6% respectively, while maintaining robust minimum improvements of 38.2%, 39.5%, and 35.2%. These results provide conclusive evidence for both the superiority of MMIGAN and the effectiveness of multimodal information fusion in addressing image inpainting challenges. The code is available at <span><span>https://github.com/guoynow/MMIGAN.git</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"712 ","pages":"Article 122159"},"PeriodicalIF":8.1,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143760782","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}

引用次数: 0

A training-free framework for valid object counting by cascading spatial and semantic understanding of foundation models

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-04-02 DOI: 10.1016/j.ins.2025.122161

Qinghong Huang , Yifan Zhang , Wenbo Zhang , Jianfeng Lin , Binqiang Huang , Jinlu Zhang , Wenhao Yu

{"title":"A training-free framework for valid object counting by cascading spatial and semantic understanding of foundation models","authors":"Qinghong Huang , Yifan Zhang , Wenbo Zhang , Jianfeng Lin , Binqiang Huang , Jinlu Zhang , Wenhao Yu","doi":"10.1016/j.ins.2025.122161","DOIUrl":"10.1016/j.ins.2025.122161","url":null,"abstract":"<div><div>Object counting aims to estimate the number of specific objects within an image. Though the estimated numeric metrics (e.g., MAE) by current methods are getting close to the ground truth, the validity and reliability of the counting objects are generally neglected. In this paper, we propose a training-free framework, called ValidCounter, to improve the validity of object counting. Two progressive modules are designed, including a spatial location module and a semantic filtering module. Given focused objects, the first module leverages the Segment Anything Model (SAM) to identify the location of objects with similar spatial features. Subsequently, the second module integrates a pre-trained vision-language encoder to compare semantic features between the focused and identified objects for filtering irrelevant ones. Our framework enhances counting accuracy by ensuring each counting object is similar to focused objects both spatially and semantically. Furthermore, as the used models are based on open vocabulary, our framework is class-agnostic and training-free, and comprehensive experiments on different datasets validate its effectiveness. Experimental results on three public datasets demonstrate that the proposed method achieves 7%-12% and 2%-6% improvement in Precision and F1 over the existing training-free approaches while preserving the competitiveness of other performance metrics. Moreover, our method achieves state-of-the-art performance on counting metrics for the FSCD-LVIS unseen-test dataset, along with top-tier effectiveness. Therefore, it can be stated that the proposed method reduces counting errors while enhancing the reliability of the counting source.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"712 ","pages":"Article 122161"},"PeriodicalIF":8.1,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143760444","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}

引用次数: 0

Iterative program synthesis with code knowledge

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-04-02 DOI: 10.1016/j.ins.2025.122156

Jiaxin Liu , Yiwei Li , Jianhua Dai , Rongjia Xu , Yating Zhang , Wei Dong

{"title":"Iterative program synthesis with code knowledge","authors":"Jiaxin Liu , Yiwei Li , Jianhua Dai , Rongjia Xu , Yating Zhang , Wei Dong","doi":"10.1016/j.ins.2025.122156","DOIUrl":"10.1016/j.ins.2025.122156","url":null,"abstract":"<div><div>Component-based synthesis aims to automatically synthesize programs from a collection of components, such as Java programs composed of application programming interfaces (APIs). However, synthesizing complex tasks remains inefficient due to the intractability of program space, which grows exponentially with the length of code. When faced with unfamiliar and complex tasks, programmers often turn to the code available online, reusing existing code knowledge to solve them. Inspired by this, we propose <span>Itas</span>, an explainable iterative program synthesis framework that incorporates code knowledge to reduce the complexity of program space and leverages modular encapsulation to synthesize more functions. Specifically, <span>Itas</span> narrows the program space externally with API knowledge, expands the program space internally with an iterative strategy to improve synthesis efficiency, and encapsulates non-API elements as general APIs to enhance modularity. To evaluate the effectiveness of <span>Itas</span>, we collect a benchmark of real programming tasks and conduct experiments on the conventional program synthesizers <span>SyPet</span> and <span>FrAngel</span>. The experimental results demonstrate that integrating with <span>Itas</span> substantially enhances synthesis efficiency, reducing the synthesis time by 97.02% compared to <span>SyPet</span> and 48.43% compared to <span>FrAngel</span>. Furthermore, comparisons with large language models show that <span>Itas</span> outperforms them in pass rate, successfully solving all tasks.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"712 ","pages":"Article 122156"},"PeriodicalIF":8.1,"publicationDate":"2025-04-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143761238","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}

引用次数: 0

Balancing the trade-off between global and personalized performance in federated learning

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-04-01 DOI: 10.1016/j.ins.2025.122154

Zibin Pan , Chi Li , Fangchen Yu , Shuyi Wang , Xiaoying Tang , Junhua Zhao

{"title":"Balancing the trade-off between global and personalized performance in federated learning","authors":"Zibin Pan , Chi Li , Fangchen Yu , Shuyi Wang , Xiaoying Tang , Junhua Zhao","doi":"10.1016/j.ins.2025.122154","DOIUrl":"10.1016/j.ins.2025.122154","url":null,"abstract":"<div><div>Balancing the trade-off between global and personalized performance has been considered a critical problem in Federated Learning (FL). On the one hand, FL using a single global model often achieves good average performance across clients but lacks personalization, resulting in poor performance on individual client's local data. On the other hand, Personalized Federated Learning (PFL) methods may fit clients well but tend to sacrifice generalization to an uncontrollable extent. To address this issue, we propose a <strong>G</strong>eneralization <strong>P</strong>reserving <strong>P</strong>ersonalized <strong>F</strong>ederated <strong>L</strong>earning algorithm (GPPFL) to achieve a robust trade-off between global and personalized performance. In GPPFL, we first enhance the average performance of the global model by formulating a multi-objective optimization problem with a fair-driven objective for FL. We calculate a common descent direction to update the global model while simultaneously mitigating the update bias and conflicts caused by absent clients. We then design a direction-drift method to identify personalized directions, building personalized models that would not sacrifice global performance. We further demonstrate how GPPFL ensures the preservation of global performance and guarantees convergence. Extensive experiments across various scenarios verify that GPPFL outperforms state-of-the-art PFL algorithms in effectively balancing the trade-off between global and personalized performance in FL.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"712 ","pages":"Article 122154"},"PeriodicalIF":8.1,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143760550","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}

引用次数: 0

A three-way efficacy evaluation approach with attribute reduction based on weighted temporal fuzzy rough sets

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-04-01 DOI: 10.1016/j.ins.2025.122157

Jin Ye , Bingzhen Sun , Xixuan Zhao , Xiaoli Chu

{"title":"A three-way efficacy evaluation approach with attribute reduction based on weighted temporal fuzzy rough sets","authors":"Jin Ye , Bingzhen Sun , Xixuan Zhao , Xiaoli Chu","doi":"10.1016/j.ins.2025.122157","DOIUrl":"10.1016/j.ins.2025.122157","url":null,"abstract":"<div><div>Precise evaluation of clinical efficacy is essential to promote the quality of treatment for patients. Accordingly, some techniques have been used to evaluate or predict the effect of treatment programs, such as statistics, machine learning, and granular computing. In contrast, granular computing can offer flexible, interpretable methods that do not require any prior knowledge to address complex efficacy evaluation problems from a data-driven perspective. Moreover, patients' efficacy information often alternates dynamically between improvement, deterioration, and no change. Granular computing-based methods provide a useful tool for the realization of three-way efficacy classification. Leveraging these advantages, this study attempts to construct a new decision-making approach over the framework of granular computing to deal with a class of efficacy evaluation problems with multi-granularity unbalanced temporal incomplete hybrid decision systems (MGUTIHDSs). To eliminate redundant attributes, we first put forward an attribute reduction method based on weighted temporal fuzzy rough sets. At the same time, several relevant properties are explored. Then, we devise a three-way efficacy evaluation model to objectively complete the personalized evaluation of previous treatment programs. Notwithstanding, it is not feasible to evaluate the efficacy of treatment programs taken at the current time node. To address this issue, a neighborhood-based average similarity prediction method is further developed. Consequently, a three-stage approach including attribute reduction, efficacy evaluation, and efficacy prediction is presented to achieve the efficacy classification of all treatment programs. Finally, the suitability and effectiveness of the approach are demonstrated by a real case study of rheumatoid arthritis. The comparison results indicate that our approach has superior performance, which can provide effective decision support for clinical practice.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"712 ","pages":"Article 122157"},"PeriodicalIF":8.1,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748158","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}

引用次数: 0

A Bagging algorithm for imprecise classification in cost-sensitive scenarios

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-04-01 DOI: 10.1016/j.ins.2025.122151

Serafín Moral-García , Andrés R. Masegosa , Joaquín Abellán

{"title":"A Bagging algorithm for imprecise classification in cost-sensitive scenarios","authors":"Serafín Moral-García , Andrés R. Masegosa , Joaquín Abellán","doi":"10.1016/j.ins.2025.122151","DOIUrl":"10.1016/j.ins.2025.122151","url":null,"abstract":"<div><div>Classic classification aims to predict the value of a variable under study, also called the class variable, based on an attribute set associated with a given item. When the evidence is not strong enough to determine such a value, a set of values of the class variable probably represents a more informative situation. This is called Imprecise Classification. An important aspect of any classification task is the cost that must be assumed in the case of error. Previous research has shown that the fusion/combination of classifiers tends to obtain better predictive results. In Imprecise Classification, there are very few models capable of efficiently fusing information from multiple classifiers. Concerning Imprecise Classification considering error costs, there is no method for this aim in the literature so far. This work presents the first method capable of fusing imprecise classifiers that take into account error costs. To do this, a procedure representing a midpoint between misclassification risk and informative outputs is used as a basis. Experiments highlight that our proposed fusion procedure shows an improvement in the results over those obtained by other methods of the state-of-the-art.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"712 ","pages":"Article 122151"},"PeriodicalIF":8.1,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143760970","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}

引用次数: 0

Maximum principle for stochastic partial differential system with fractional Brownian motion

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-04-01 DOI: 10.1016/j.ins.2025.122144

Xiaolin Yuan , Guojian Ren , YangQuan Chen , Yongguang Yu

引用次数: 0

Optimized consensus control of multi-manipulator system having actuator fault using reinforcement learning approximation strategy

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-03-31 DOI: 10.1016/j.ins.2025.122141

Yu Cao , Guoxing Wen , Baoshuo Feng , Bin Li

{"title":"Optimized consensus control of multi-manipulator system having actuator fault using reinforcement learning approximation strategy","authors":"Yu Cao , Guoxing Wen , Baoshuo Feng , Bin Li","doi":"10.1016/j.ins.2025.122141","DOIUrl":"10.1016/j.ins.2025.122141","url":null,"abstract":"<div><div>This work is to develop an optimized consensus control of multi-manipulator system by employing reinforcement learning (RL) approximation strategy, while the multi-manipulator system is supposed to have the problem of actuator uncertain fault. The RL approximation strategy aims to avoid directly solving the Hamilton-Jacobi-Bellman (HJB) equation for finding the optimized consensus control because this equation has the strong nonlinearity. Since the manipulator system is modeled by the double-integral dynamic, the RL algorithm needs to simultaneously involve two position and velocity states. Meanwhile, it needs to also consider the compensation of actuator fault consisting of both time-varying efficiency factor and bias signal. By defining the performance function containing the above information, the optimized leader-follower consensus can be competent to work for the multi-manipulator system having actuator fault. Finally, the feasibility and validation of this optimizing consensus method is demonstrated from the two aspects of theory analysis and computer simulation.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"712 ","pages":"Article 122141"},"PeriodicalIF":8.1,"publicationDate":"2025-03-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143760971","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}

引用次数: 0

Zilean: A modularized framework for large-scale temporal concept drift type classification

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-03-28 DOI: 10.1016/j.ins.2025.122134

Zhao Deng , QuanXi Feng , Bin Lin , Gary G. Yen

{"title":"Zilean: A modularized framework for large-scale temporal concept drift type classification","authors":"Zhao Deng , QuanXi Feng , Bin Lin , Gary G. Yen","doi":"10.1016/j.ins.2025.122134","DOIUrl":"10.1016/j.ins.2025.122134","url":null,"abstract":"<div><div>In the analysis of time series data, particularly in real-world applications, concept drift classification is crucial for enabling models to adapt in a differentiated manner to future data. To address the challenge of identifying diverse types of drift, we propose Zilean, a novel framework that integrates feature-based and predictor-based techniques while accounting for drift residues and fragmentation during repeated drift detection. The framework incorporates the pre-trained BERT-Base language model into its classifier design, leveraging deep learning for automatic drift classification and eliminating the need for judgment curve analysis. To evaluate its performance, experiments were conducted on a variety of real-world and synthetic datasets, each exhibiting different types of concept drift. The results show that on real-world datasets, our framework achieves a classification accuracy of 91.03%, outperforming XGBoost by 7.94% and surpassing TCN-CNN by 4.28%. Additionally, experiments exploring a frozen parameter strategy and the use of a more lightweight language model, DistilBERT, further enhance accuracy to 96.93% and 97.17%, respectively. These findings underscore the framework's effectiveness in large-scale temporal concept drift classification.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"712 ","pages":"Article 122134"},"PeriodicalIF":8.1,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143739395","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}

引用次数: 0

A novel missing value imputation for multi-criteria recommender systems

IF 8.1 1区计算机科学

Information Sciences Pub Date : 2025-03-28 DOI: 10.1016/j.ins.2025.122139

Gökhan Göksel, Ahmet Aydın, Zeynep Batmaz, Cihan Kaleli

{"title":"A novel missing value imputation for multi-criteria recommender systems","authors":"Gökhan Göksel, Ahmet Aydın, Zeynep Batmaz, Cihan Kaleli","doi":"10.1016/j.ins.2025.122139","DOIUrl":"10.1016/j.ins.2025.122139","url":null,"abstract":"<div><div>As the internet continues to expand, users increasingly rely on digital platforms for activities such as movie streaming and travel planning, leading to an overwhelming amount of information. This information overload complicates the process of making efficient and informed decisions. To address this challenge, personalized recommender systems, particularly multi-criteria collaborative filtering (MCCF) models, have been developed to tailor recommendations based on detailed user evaluations across various sub-criteria. However, as the number of criteria in MCCF systems increases, the issue of data sparsity becomes more prominent, with more criteria resulting in more missing evaluations. In this study, we introduce an innovative application of the Bidirectional Encoder Representations from Transformers (BERT) model—well-known for its advances in natural language processing—to impute missing values within MCCF systems. By leveraging contextual insights from user reviews, we hypothesize that BERT can enhance the imputation process, thereby improving coverage and recommendation accuracy in MCCF models. Our experimental results indicate that BERT-based imputation significantly reduces data sparsity and enhances the accuracy and coverage of recommendations. This study underscores BERT's potential in processing linguistic data and highlights its utility in multi-criteria recommender systems. Integrating BERT with MCCF offers a promising advancement in addressing the inherent challenges of personalized recommendation systems.</div></div>","PeriodicalId":51063,"journal":{"name":"Information Sciences","volume":"712 ","pages":"Article 122139"},"PeriodicalIF":8.1,"publicationDate":"2025-03-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143739391","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}

引用次数: 0