Knowledge-Based Systems最新文献_第3页

Capsule based regressor network for multivariate short term weather forecasting 基于胶囊的多元短期天气预报回归网络

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-05 DOI: 10.1016/j.knosys.2026.115439

Arjun Mallick, Arkadeep De, Arpan Nandi, Asif Iqbal Middya, Sarbani Roy

The problem of weather forecasting has always been in the forefront of studies regarding time series analysis and has a long trailing history of applied techniques due to its immense importance in science as well as our socio-economic lives. In recent years, short-term weather forecasting has evolved rapidly, driven by availability of enormous amounts of data, exponential lift in computation feasibility, and theoretical progress in machine learning. The paradigm of weather forecasting is gradually shifting from simulation-based physics modeling methods to more data centric methods - resulting in more accurate and real time forecasts. This study frames the weather forecasting problem as a multivariate time series problem specific to a fixed location and introduces the method of dynamic routing between capsules to the weather forecasting paradigm. Learning from some selected parameters and their inter dependencies, the capsule regressor network forecasts the temperature, humidity, wind speed, sea level pressure and vapor pressure for next timesteps. We have rigorously compared its performance against all broader varieties of neural networks which have been applied in weather forecasting - and it was observed that the capsule regressor worked fairly well within the forecast horizon of 120 h. It outperformed all other baselines in 48 h and 72 h forecast horizons and remained close to best in other timesteps. The study also portrays a measure of the genericness of models predicting different features with unique characteristics and across all horizons, where the capsule network was found to be the most consistent.

天气预报问题一直处于时间序列分析研究的前沿，由于其在科学以及我们的社会经济生活中的巨大重要性，它具有悠久的应用技术历史。近年来，由于大量数据的可用性、计算可行性的指数提升以及机器学习的理论进步，短期天气预报发展迅速。天气预报的模式正逐渐从基于仿真的物理建模方法转向以数据为中心的方法，从而实现更准确、更实时的预报。本研究将天气预报问题作为一个特定于固定位置的多变量时间序列问题，并将胶囊之间的动态路由方法引入天气预报范式。从一些选定的参数及其相互依赖关系中学习，胶囊回归网络预测下一个时间步的温度、湿度、风速、海平面压力和蒸汽压。我们将其性能与应用于天气预报的所有更广泛的神经网络进行了严格的比较，并观察到胶囊回归器在120小时的预测范围内工作得相当好。在48小时和72小时的预测范围内，它的表现优于所有其他基线，并在其他时间步长中保持接近最佳。该研究还描绘了一种模型的通用性，该模型预测了具有独特特征的不同特征，并跨越了所有的视野，其中胶囊网络被发现是最一致的。

{"title":"Capsule based regressor network for multivariate short term weather forecasting","authors":"Arjun Mallick, Arkadeep De, Arpan Nandi, Asif Iqbal Middya, Sarbani Roy","doi":"10.1016/j.knosys.2026.115439","DOIUrl":"10.1016/j.knosys.2026.115439","url":null,"abstract":"<div><div>The problem of weather forecasting has always been in the forefront of studies regarding time series analysis and has a long trailing history of applied techniques due to its immense importance in science as well as our socio-economic lives. In recent years, short-term weather forecasting has evolved rapidly, driven by availability of enormous amounts of data, exponential lift in computation feasibility, and theoretical progress in machine learning. The paradigm of weather forecasting is gradually shifting from simulation-based physics modeling methods to more data centric methods - resulting in more accurate and real time forecasts. This study frames the weather forecasting problem as a multivariate time series problem specific to a fixed location and introduces the method of dynamic routing between capsules to the weather forecasting paradigm. Learning from some selected parameters and their inter dependencies, the capsule regressor network forecasts the temperature, humidity, wind speed, sea level pressure and vapor pressure for next timesteps. We have rigorously compared its performance against all broader varieties of neural networks which have been applied in weather forecasting - and it was observed that the capsule regressor worked fairly well within the forecast horizon of 120 h. It outperformed all other baselines in 48 h and 72 h forecast horizons and remained close to best in other timesteps. The study also portrays a measure of the genericness of models predicting different features with unique characteristics and across all horizons, where the capsule network was found to be the most consistent.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"338 ","pages":"Article 115439"},"PeriodicalIF":7.6,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174782","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

Confusion-Calibrated Cross-Entropy and Class-Specialized Aggregation for Robust Federated Learning under Extreme Data Heterogeneity 基于混淆校正的交叉熵和类专用聚合的极端数据异构鲁棒联邦学习

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-05 DOI: 10.1016/j.knosys.2026.115497

Sujit Chowdhury , Raju Halder

Federated Learning (FL) enables privacy-preserving collaborative model training across decentralized devices without exchanging raw data. However, its effectiveness is severely hampered by data heterogeneity – particularly label distribution skew, missing classes, and data sparsity-which causes model divergence, poor generalization, and unfair performance across classes. Traditional FL methods rely on uniform aggregation and standard loss functions that fail to account for local biases and class-level struggles, leading to catastrophic degradation under realistic data heterogeneity conditions. To this end, we introduce FedCA, a federated learning framework that jointly mitigates label skew, missing classes, and data sparsity through two synergistic components: a Confusion-Calibrated Cross-Entropy (C3E) loss for client-side training and a struggling-class-prioritized top-kaggregation scheme at the server. C3E dynamically calibrates local objectives using an adaptive, on-device soft confusion matrix that penalizes persistent misclassifications to correct client-level biases. Complementing this, the aggregation module leverages a compact struggler signal from clients to prioritize updates targeting the most challenging classes across the federation, enhancing both robustness and fairness. Extensive experiments on MNIST, CIFAR-10, and CIFAR-100 demonstrate that FedCA significantly outperforms state-of-the-art baselines. Under extreme label distribution skew, FedCA achieves a 20.12% absolute accuracy advantage on CIFAR-10. It maintains 63.1% accuracy when clients hold only two classes-a regime where baselines fail-and reaches target accuracies in up to 4 × fewer rounds. These results validate FedCA as a statistically grounded, communication-efficient, and robust solution for high-performance federated learning under realistic data heterogeneity conditions.

联邦学习（FL）可以在不交换原始数据的情况下跨分散设备进行保护隐私的协作模型训练。然而，它的有效性受到数据异质性的严重阻碍——特别是标签分布倾斜、缺失类和数据稀疏性——这会导致模型分歧、泛化不良和跨类的不公平性能。传统的FL方法依赖于统一的聚合和标准损失函数，这些函数不能考虑局部偏差和类水平的斗争，导致在实际数据异质性条件下的灾难性退化。为此，我们引入了FedCA，这是一个联邦学习框架，它通过两个协同组件共同减轻了标签倾斜、缺失类和数据稀疏性：用于客户端训练的混淆校准交叉熵（C3E）损失和服务器端的挣扎类优先级顶级聚合方案。C3E使用自适应的、设备上的软混淆矩阵来动态校准本地目标，该矩阵会惩罚持续的错误分类，以纠正客户级偏差。与此相补充的是，聚合模块利用来自客户端的紧凑的挣扎者信号来优先考虑针对整个联邦中最具挑战性的类的更新，从而增强了鲁棒性和公平性。在MNIST、CIFAR-10和CIFAR-100上进行的大量实验表明，FedCA的性能明显优于最先进的基线。在极端标签分布偏斜的情况下，FedCA在CIFAR-10上获得了20.12%的绝对精度优势。当客户端只持有两个类别（基线失效的情况）时，它保持63.1%的准确性，并在最多4次 × 少轮中达到目标准确性。这些结果验证了FedCA是一种基于统计的、通信高效的、健壮的解决方案，适用于实际数据异构条件下的高性能联邦学习。

{"title":"Confusion-Calibrated Cross-Entropy and Class-Specialized Aggregation for Robust Federated Learning under Extreme Data Heterogeneity","authors":"Sujit Chowdhury , Raju Halder","doi":"10.1016/j.knosys.2026.115497","DOIUrl":"10.1016/j.knosys.2026.115497","url":null,"abstract":"<div><div>Federated Learning (FL) enables privacy-preserving collaborative model training across decentralized devices without exchanging raw data. However, its effectiveness is severely hampered by data heterogeneity – particularly label distribution skew, missing classes, and data sparsity-which causes model divergence, poor generalization, and unfair performance across classes. Traditional FL methods rely on uniform aggregation and standard loss functions that fail to account for local biases and class-level struggles, leading to catastrophic degradation under realistic data heterogeneity conditions. To this end, we introduce <span>FedCA</span>, a federated learning framework that jointly mitigates label skew, missing classes, and data sparsity through two synergistic components: a <strong>Confusion-Calibrated Cross-Entropy (C3E)</strong> loss for client-side training and a <strong>struggling-class-prioritized top-</strong><em>k</em><strong>aggregation</strong> scheme at the server. C3E dynamically calibrates local objectives using an adaptive, on-device soft confusion matrix that penalizes persistent misclassifications to correct client-level biases. Complementing this, the aggregation module leverages a compact struggler signal from clients to prioritize updates targeting the most challenging classes across the federation, enhancing both robustness and fairness. Extensive experiments on MNIST, CIFAR-10, and CIFAR-100 demonstrate that <span>FedCA</span> significantly outperforms state-of-the-art baselines. Under extreme label distribution skew, <span>FedCA</span> achieves a <strong>20.12%</strong> absolute accuracy advantage on CIFAR-10. It maintains <strong>63.1%</strong> accuracy when clients hold only two classes-a regime where baselines fail-and reaches target accuracies in up to <strong>4</strong> × <strong>fewer rounds</strong>. These results validate <span>FedCA</span> as a statistically grounded, communication-efficient, and robust solution for high-performance federated learning under realistic data heterogeneity conditions.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"338 ","pages":"Article 115497"},"PeriodicalIF":7.6,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175183","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 frequency and spatial domain interactive fusion network for underwater image enhancement 一种用于水下图像增强的频域与空域交互融合网络

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-05 DOI: 10.1016/j.knosys.2026.115499

Guangfeng Li, Shiming Sun, Chengwen Zhang, Guangsen Jiao

Underwater images often suffer from degradation such as color distortion and detail blurring due to light absorption and scattering. Despite numerous underwater image enhancement methods have been proposed to improve visual quality, most of them are limited to processing spatial domain features and neglect the inherent frequency domain information, thereby affecting the enhancement performance. To address this issue, we propose a Frequency and Spatial domain Interactive Fusion Network (FSIF-Net) for underwater image enhancement. Specifically, we first design a Frequency Domain Enhancement Module (FDEM). This module leverages the feature decomposition capability of wavelet transform alongside the frequency domain modeling advantages of the Fast Fourier Transform (FFT), effectively enhancing global color while improving edge and texture information. Secondly, we design a Local Detail Enhancement Module (LDEM), which utilizes horizontal, vertical, and diagonal convolutional operations to enhance anisotropic image features and introduces a sliding window mechanism to improve local detail enhancement capability. Finally, to achieve complementary fusion of frequency and spatial domain features, we design a Dual-domain Interactive Fusion Module (DIFM). This module adaptively acquires more representative frequency and spatial features through a weight-reshaping gating mechanism, followed by comprehensive fusion of the dual-domain features across both channel and spatial dimensions. Extensive experiments demonstrate that the proposed FSIF-Net significantly enhances the visual quality of underwater images and outperforms state-of-the-art methods in both quantitative and qualitative evaluations.

由于光的吸收和散射，水下图像经常遭受诸如颜色失真和细节模糊等退化。为了提高水下图像的视觉质量，人们提出了许多水下图像增强方法，但大多数方法都局限于处理空间域特征，忽略了固有的频域信息，从而影响了增强效果。为了解决这个问题，我们提出了一种用于水下图像增强的频率和空间域交互融合网络（FSIF-Net）。具体来说，我们首先设计了一个频域增强模块（FDEM）。该模块利用小波变换的特征分解能力和快速傅里叶变换（FFT）的频域建模优势，在提高边缘和纹理信息的同时，有效增强了图像的全局色彩。其次，我们设计了局部细节增强模块（LDEM），利用水平、垂直和对角卷积运算增强图像的各向异性特征，并引入滑动窗口机制来提高局部细节增强能力。最后，为了实现频率域和空间域特征的互补融合，设计了双域交互融合模块（DIFM）。该模块通过权重重塑门控机制自适应获取更具代表性的频率和空间特征，然后在通道和空间维度上对双域特征进行全面融合。大量的实验表明，所提出的FSIF-Net显著提高了水下图像的视觉质量，在定量和定性评估方面都优于最先进的方法。

{"title":"A frequency and spatial domain interactive fusion network for underwater image enhancement","authors":"Guangfeng Li, Shiming Sun, Chengwen Zhang, Guangsen Jiao","doi":"10.1016/j.knosys.2026.115499","DOIUrl":"10.1016/j.knosys.2026.115499","url":null,"abstract":"<div><div>Underwater images often suffer from degradation such as color distortion and detail blurring due to light absorption and scattering. Despite numerous underwater image enhancement methods have been proposed to improve visual quality, most of them are limited to processing spatial domain features and neglect the inherent frequency domain information, thereby affecting the enhancement performance. To address this issue, we propose a Frequency and Spatial domain Interactive Fusion Network (FSIF-Net) for underwater image enhancement. Specifically, we first design a Frequency Domain Enhancement Module (FDEM). This module leverages the feature decomposition capability of wavelet transform alongside the frequency domain modeling advantages of the Fast Fourier Transform (FFT), effectively enhancing global color while improving edge and texture information. Secondly, we design a Local Detail Enhancement Module (LDEM), which utilizes horizontal, vertical, and diagonal convolutional operations to enhance anisotropic image features and introduces a sliding window mechanism to improve local detail enhancement capability. Finally, to achieve complementary fusion of frequency and spatial domain features, we design a Dual-domain Interactive Fusion Module (DIFM). This module adaptively acquires more representative frequency and spatial features through a weight-reshaping gating mechanism, followed by comprehensive fusion of the dual-domain features across both channel and spatial dimensions. Extensive experiments demonstrate that the proposed FSIF-Net significantly enhances the visual quality of underwater images and outperforms state-of-the-art methods in both quantitative and qualitative evaluations.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"338 ","pages":"Article 115499"},"PeriodicalIF":7.6,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146174778","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

SGC: A self-guided cascade multitask model for low-light object detection 弱光目标检测的自引导级联多任务模型

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-05 DOI: 10.1016/j.knosys.2026.115462

Jiakun Jin , Junchao Zhang , Yidong Luo , Jiandong Tian

Outdoor scenes often suffer from insufficient and non-uniform illumination, leading to object detection (OD) failures. This issue has garnered research attention, with the mainstream solution being to improve the model’s feature extraction capability through cascaded feature enhancement modules. However, such approaches increase the model’s complexity and the enhancement effect is highly dependent on the similarity between the training and testing data. Alternatively, some methods incorporate parallel low-light image enhancement (LLE) modules to guide the training of object detection models. Nevertheless, due to the lack of object detection datasets containing paired bright and low-light images, these methods often require manually selecting appropriate pre-trained LLE models for different scenes, making end-to-end training challenging. In this paper, we aim to build an end-to-end LLE&OD cascade multitask model that leverages the strengths of both approaches. We use a new data augmentation techniques to synthesize low-light images from normal-light object detection datasets. To mutually train the cascade model, a new self-guided loss is designed. By deconstruction and reorganization of the multitask model, the self-guided loss effectively steering the model away from local optima for single tasks, enabling the model to achieve superior performance compared to many state-of-the-art methods on several publicly available night scene datasets, as well as on a daytime scene dataset. The source code of the proposed method will be available at https://github.com/225ceV/SGC.

户外场景经常受到光照不足和不均匀的影响，导致物体检测（OD）失败。这个问题已经引起了研究的关注，主流的解决方案是通过级联的特征增强模块来提高模型的特征提取能力。然而，这种方法增加了模型的复杂性，并且增强效果高度依赖于训练数据和测试数据之间的相似度。或者，一些方法结合并行低光图像增强（LLE）模块来指导目标检测模型的训练。然而，由于缺乏包含成对的明亮和低光图像的目标检测数据集，这些方法通常需要为不同的场景手动选择适当的预训练LLE模型，这使得端到端训练具有挑战性。在本文中，我们的目标是建立一个端到端的lleod级联多任务模型，利用这两种方法的优势。我们使用一种新的数据增强技术，从正常光目标检测数据集合成低光图像。为了相互训练串级模型，设计了一种新的自导向损失。通过对多任务模型的解构和重组，自引导损失有效地使模型远离单个任务的局部最优，使模型在几个公开可用的夜景数据集以及白天场景数据集上获得比许多最先进的方法更好的性能。建议的方法的源代码可以在https://github.com/225ceV/SGC上获得。

{"title":"SGC: A self-guided cascade multitask model for low-light object detection","authors":"Jiakun Jin , Junchao Zhang , Yidong Luo , Jiandong Tian","doi":"10.1016/j.knosys.2026.115462","DOIUrl":"10.1016/j.knosys.2026.115462","url":null,"abstract":"<div><div>Outdoor scenes often suffer from insufficient and non-uniform illumination, leading to object detection (OD) failures. This issue has garnered research attention, with the mainstream solution being to improve the model’s feature extraction capability through cascaded feature enhancement modules. However, such approaches increase the model’s complexity and the enhancement effect is highly dependent on the similarity between the training and testing data. Alternatively, some methods incorporate parallel low-light image enhancement (LLE) modules to guide the training of object detection models. Nevertheless, due to the lack of object detection datasets containing paired bright and low-light images, these methods often require manually selecting appropriate pre-trained LLE models for different scenes, making end-to-end training challenging. In this paper, we aim to build an end-to-end LLE&OD cascade multitask model that leverages the strengths of both approaches. We use a new data augmentation techniques to synthesize low-light images from normal-light object detection datasets. To mutually train the cascade model, a new self-guided loss is designed. By deconstruction and reorganization of the multitask model, the self-guided loss effectively steering the model away from local optima for single tasks, enabling the model to achieve superior performance compared to many state-of-the-art methods on several publicly available night scene datasets, as well as on a daytime scene dataset. The source code of the proposed method will be available at <span><span>https://github.com/225ceV/SGC</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"338 ","pages":"Article 115462"},"PeriodicalIF":7.6,"publicationDate":"2026-02-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175314","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

LLMs For drug-Drug interaction prediction using textual drug descriptors 使用文本药物描述符进行药物-药物相互作用预测

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-04 DOI: 10.1016/j.knosys.2026.115486

Gabriele De Vito , Filomena Ferrucci , Athanasios Angelakis

As treatment plans involve more medications, anticipating and preventing drug-drug interactions (DDIs) becomes increasingly important. Such interactions can result in harmful side effects and may reduce therapy effectiveness. Currently, most computational approaches for DDI prediction rely heavily on complex feature engineering and require chemical information to be structured in specific formats to enable accurate detection of potential interactions. This study presents the first investigation of the application of Large Language Models (LLMs) for DDI prediction using drug characteristics expressed solely in free-text form. Specifically, we use SMILES notations, target organisms, and gene associations as inputs in purpose-designed prompts, allowing LLMs to learn the underlying relationships among these descriptors and accordingly predict possible DDIs. We evaluated the performance of 18 distinct LLMs under zero-shot, few-shot, and fine-tuning settings on the DrugBank dataset (version 5.1.12) to identify the most effective paradigm. We then assessed the generalizability of the fine-tuned models on 13 external DDI datasets against well-known machine learning baselines. The results demonstrated that, while zero-shot and few-shot paradigms showed only modest utility, fine-tuned models achieved superior sensitivity while maintaining competitive accuracy and F1-score compared to baselines. Notably, despite its small size, the Phi-3.5 2.7B model attained a sensitivity of 0.978 and an accuracy of 0.919. These findings suggest that computational efficiency and task-specific adaptation are more important than model size in order to capture the complex patterns inherent in drug interactions, and outline a more accessible paradigm for DDI prediction that can be integrated into clinical decision support systems.

随着治疗方案涉及更多的药物，预测和预防药物相互作用（ddi）变得越来越重要。这种相互作用可能导致有害的副作用，并可能降低治疗效果。目前，大多数用于DDI预测的计算方法严重依赖于复杂的特征工程，并且需要以特定格式构建化学信息，以便准确检测潜在的相互作用。本研究首次研究了大型语言模型（LLMs）在DDI预测中的应用，该模型使用仅以自由文本形式表达的药物特征。具体来说，我们使用SMILES符号、目标生物和基因关联作为目的设计提示符的输入，允许llm学习这些描述符之间的潜在关系，并相应地预测可能的ddi。我们在DrugBank数据集（版本5.1.12）上评估了18种不同llm在零射击、少射击和微调设置下的性能，以确定最有效的范式。然后，我们根据众所周知的机器学习基线，在13个外部DDI数据集上评估了微调模型的泛化性。结果表明，虽然零射击和少射击范式仅显示适度的效用，但微调模型在保持竞争精度和f1分数的同时获得了更高的灵敏度。值得注意的是，尽管尺寸较小，但Phi-3.5 2.7B模型的灵敏度为0.978，精度为0.919。这些发现表明，为了捕捉药物相互作用中固有的复杂模式，计算效率和任务特异性适应比模型大小更重要，并概述了可集成到临床决策支持系统的DDI预测更容易获得的范式。

{"title":"LLMs For drug-Drug interaction prediction using textual drug descriptors","authors":"Gabriele De Vito , Filomena Ferrucci , Athanasios Angelakis","doi":"10.1016/j.knosys.2026.115486","DOIUrl":"10.1016/j.knosys.2026.115486","url":null,"abstract":"<div><div>As treatment plans involve more medications, anticipating and preventing drug-drug interactions (DDIs) becomes increasingly important. Such interactions can result in harmful side effects and may reduce therapy effectiveness. Currently, most computational approaches for DDI prediction rely heavily on complex feature engineering and require chemical information to be structured in specific formats to enable accurate detection of potential interactions. This study presents the first investigation of the application of Large Language Models (LLMs) for DDI prediction using drug characteristics expressed solely in free-text form. Specifically, we use SMILES notations, target organisms, and gene associations as inputs in purpose-designed prompts, allowing LLMs to learn the underlying relationships among these descriptors and accordingly predict possible DDIs. We evaluated the performance of 18 distinct LLMs under zero-shot, few-shot, and fine-tuning settings on the DrugBank dataset (version 5.1.12) to identify the most effective paradigm. We then assessed the generalizability of the fine-tuned models on 13 external DDI datasets against well-known machine learning baselines. The results demonstrated that, while zero-shot and few-shot paradigms showed only modest utility, fine-tuned models achieved superior sensitivity while maintaining competitive accuracy and F1-score compared to baselines. Notably, despite its small size, the Phi-3.5 2.7B model attained a sensitivity of 0.978 and an accuracy of 0.919. These findings suggest that computational efficiency and task-specific adaptation are more important than model size in order to capture the complex patterns inherent in drug interactions, and outline a more accessible paradigm for DDI prediction that can be integrated into clinical decision support systems.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"338 ","pages":"Article 115486"},"PeriodicalIF":7.6,"publicationDate":"2026-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175322","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

ATARS: Adaptive task-Aware feature learning for Few-Shot Fine-Grained classification ATARS：自适应任务感知特征学习，用于少量细粒度分类

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-04 DOI: 10.1016/j.knosys.2026.115485

Xiaomei Long, Xinyue Wang, Cheng Yang, Zongbo He, Qian He, Xiangdong Chen

Few-shot fine-grained classification is challenging due to subtle inter-class differences and limited annotations. Existing methods often fail to fully exploit task-level information, limiting adaptation to scarce samples. We present ATARS, a task-aware framework that organizes alignment, feature reconstruction, and task-conditioned channel selection into a coordinated pipeline. These components progressively refine task-adaptive feature representations, enhancing intra-class consistency and discriminative capacity. Extensive experiments on five fine-grained benchmarks demonstrate the effectiveness of this design: ATARS achieves 5-way 5-shot accuracies of 97.38% on Cars, 94.40% on CUB, and 89.78% on Dogs, consistently outperforming previous reconstruction-based and task-aware approaches. The results highlight the benefits of coordinated component design under task-aware guidance in few-shot scenarios. The source code is available at: https://github.com/lxm-hjk/ATARS-FSL.

由于微妙的类间差异和有限的注释，少量的细粒度分类是具有挑战性的。现有的方法往往不能充分利用任务级信息，限制了对稀缺样本的适应。我们提出了ATARS，一个任务感知框架，将对齐、特征重建和任务条件通道选择组织到一个协调的管道中。这些组件逐步细化任务自适应特征表示，增强类内一致性和判别能力。在五个细粒度基准测试上进行的大量实验证明了这种设计的有效性：ATARS在Cars上实现了97.38%的5-way 5-shot准确率，在CUB上实现了94.40%，在Dogs上实现了89.78%，始终优于以前基于重建和任务感知的方法。结果强调了在任务感知指导下的协同组件设计在少数射击场景下的好处。源代码可从https://github.com/lxm-hjk/ATARS-FSL获得。

引用次数: 0

Towards robust and high-capacity coverless image steganography 迈向鲁棒性和高容量无覆盖图像隐写

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-04 DOI: 10.1016/j.knosys.2026.115472

Bobiao Guo, Ping Ping

Capacity and robustness are key metrics for Coverless Image Steganography (CIS). Although the theoretical maximum capacity of the CIS has been achieved, the robustness at this level remains insufficient, limiting the practical use of high-capacity settings. Therefore, this paper proposes a novel CIS method that maintains high robustness even at the theoretical maximum capacity supported by a given dataset. Our method has three main parts: constructing a stable feature space using Pseudo-Zernike Moments (PZM), proposing a Stability-Aware Piecewise Quantization Encoding (SA-PQE) to assign a stability score to each image, and introducing stability regularization into the clustering process to build a robust Coverless Index Database (CID). Its robust high-capacity performance derives from two core principles: (1) low-order PZM coefficients remain highly stable under distortion, and the independence among PZM coefficients suppresses distortion propagation across dimensions; and (2) stability regularization penalizes images that are close to the cluster center but exhibit low stability. Extensive experimental results demonstrate that our method achieves high robustness at the theoretical maximum capacity. On the Holidays, VOC, and ImageNet datasets, the average robustness reaches 99.54%, 98.64%, and 97.19% at capacities of 10 bits, 14 bits, and 15 bits, respectively. These results significantly outperform existing advanced methods under the setting without inverse image retrieval.

容量和鲁棒性是无覆盖图像隐写（CIS）的关键指标。虽然独联体的理论最大容量已经达到，但这一级的稳健性仍然不够，限制了实际使用高容量设置。因此，本文提出了一种新颖的CIS方法，即使在给定数据集支持的理论最大容量下也能保持高鲁棒性。该方法主要包括三个部分：利用伪泽尼克矩（Pseudo-Zernike Moments， PZM）构建稳定的特征空间；提出一种稳定性感知的分段量化编码（SA-PQE），为每张图像分配稳定性评分；在聚类过程中引入稳定性正则化，构建鲁棒的无覆盖索引数据库（Coverless Index Database， CID）。其强大的高容量性能源于两个核心原则：(1)低阶PZM系数在失真情况下保持高度稳定，PZM系数之间的独立性抑制了失真在维度上的传播；(2)稳定性正则化对靠近聚类中心但稳定性较低的图像进行惩罚。大量的实验结果表明，该方法在理论最大容量下具有较高的鲁棒性。在节假日、VOC和ImageNet数据集上，10比特、14比特和15比特容量下的平均鲁棒性分别达到99.54%、98.64%和97.19%。在没有逆图像检索的情况下，这些结果明显优于现有的先进方法。

{"title":"Towards robust and high-capacity coverless image steganography","authors":"Bobiao Guo, Ping Ping","doi":"10.1016/j.knosys.2026.115472","DOIUrl":"10.1016/j.knosys.2026.115472","url":null,"abstract":"<div><div>Capacity and robustness are key metrics for Coverless Image Steganography (CIS). Although the theoretical maximum capacity of the CIS has been achieved, the robustness at this level remains insufficient, limiting the practical use of high-capacity settings. Therefore, this paper proposes a novel CIS method that maintains high robustness even at the theoretical maximum capacity supported by a given dataset. Our method has three main parts: constructing a stable feature space using Pseudo-Zernike Moments (PZM), proposing a Stability-Aware Piecewise Quantization Encoding (SA-PQE) to assign a stability score to each image, and introducing stability regularization into the clustering process to build a robust Coverless Index Database (CID). Its robust high-capacity performance derives from two core principles: (1) low-order PZM coefficients remain highly stable under distortion, and the independence among PZM coefficients suppresses distortion propagation across dimensions; and (2) stability regularization penalizes images that are close to the cluster center but exhibit low stability. Extensive experimental results demonstrate that our method achieves high robustness at the theoretical maximum capacity. On the Holidays, VOC, and ImageNet datasets, the average robustness reaches 99.54%, 98.64%, and 97.19% at capacities of 10 bits, 14 bits, and 15 bits, respectively. These results significantly outperform existing advanced methods under the setting without inverse image retrieval.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"338 ","pages":"Article 115472"},"PeriodicalIF":7.6,"publicationDate":"2026-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175317","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

MM-AttacKG: A multimodal approach to attack graph construction with large language models MM-AttacKG：使用大型语言模型构建攻击图的多模态方法

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-04 DOI: 10.1016/j.knosys.2026.115483

Yongheng Zhang , Xinyun Zhao , Yunshan Ma , Haokai Ma , Yingxiao Guan , Guozheng Yang , Yuliang Lu , Xiang Wang

Cyber Threat Intelligence (CTI) parsing aims to extract key threat information from massive data, transform it into actionable intelligence, enhance threat detection and defense efficiency, including attack graph construction, intelligence fusion, and indicator extraction. Among these research topics, Attack Graph Construction (AGC) is essential for visualizing and understanding the potential attack paths of threat events from CTI reports. Existing approaches primarily construct the attack graphs purely from the textual data to reveal the logical threat relationships between entities within the attack behavioral sequence. However, they typically overlook the specific threat information inherent in visual modalities, which preserves key threat details from inherently multimodal CTI reports. Inspired by the remarkable multimodal understanding capabilities of Multimodal Large Language Models (MLLMs), we explore their potential in enhancing multimodal attack graph construction. To be specific, we propose a novel framework, MM-AttacKG, which can effectively extract key information from threat images and integrate it into attack graph construction, thereby enhancing the comprehensiveness and accuracy of attack graphs. It first employs a threat image parsing module to extract critical threat information from images and generate textual descriptions using MLLMs. Subsequently, it builds an iterative question-answering pipeline tailored for image parsing to refine the understanding of threat images. Finally, it achieves content-level integration between attack graphs and image-based answers through MLLMs, completing threat information enhancement. We construct a new multimodal dataset, AG-LLM-mm, and conduct extensive experiments to evaluate the effectiveness of MM-AttacKG. The results demonstrate that MM-AttacKG can accurately identify key information in threat images and significantly improve the quality of multimodal attack graph construction, effectively addressing the shortcomings of existing methods in utilizing image-based threat information.

CTI （Cyber Threat Intelligence）分析旨在从海量数据中提取关键威胁信息，转化为可操作的情报，提高威胁检测和防御效率，包括构建攻击图、融合情报、提取指标等。在这些研究课题中，攻击图构建（AGC）对于可视化和理解CTI报告中威胁事件的潜在攻击路径至关重要。现有的方法主要是从纯文本数据构建攻击图，以揭示攻击行为序列中实体之间的逻辑威胁关系。然而，它们通常忽略了视觉模态中固有的特定威胁信息，而视觉模态保留了固有多模态CTI报告中的关键威胁细节。受多模态大型语言模型（mllm）出色的多模态理解能力的启发，我们探索了它们在增强多模态攻击图构建方面的潜力。具体而言，我们提出了一种新的框架MM-AttacKG，它可以有效地从威胁图像中提取关键信息，并将其整合到攻击图的构建中，从而提高攻击图的全能性和准确性。首先利用威胁图像解析模块，从图像中提取关键威胁信息，并利用mllm生成文本描述；随后，构建了针对图像解析定制的迭代问答管道，以细化对威胁图像的理解。最后，通过mllm实现攻击图与基于图像的答案之间的内容级集成，完成威胁信息增强。我们构建了一个新的多模态数据集AG-LLM-mm，并进行了大量的实验来评估MM-AttacKG的有效性。结果表明，MM-AttacKG能够准确识别威胁图像中的关键信息，显著提高多模态攻击图构建质量，有效解决了现有方法在利用基于图像的威胁信息方面存在的不足。

{"title":"MM-AttacKG: A multimodal approach to attack graph construction with large language models","authors":"Yongheng Zhang , Xinyun Zhao , Yunshan Ma , Haokai Ma , Yingxiao Guan , Guozheng Yang , Yuliang Lu , Xiang Wang","doi":"10.1016/j.knosys.2026.115483","DOIUrl":"10.1016/j.knosys.2026.115483","url":null,"abstract":"<div><div>Cyber Threat Intelligence (CTI) parsing aims to extract key threat information from massive data, transform it into actionable intelligence, enhance threat detection and defense efficiency, including attack graph construction, intelligence fusion, and indicator extraction. Among these research topics, Attack Graph Construction (AGC) is essential for visualizing and understanding the potential attack paths of threat events from CTI reports. Existing approaches primarily construct the attack graphs purely from the textual data to reveal the logical threat relationships between entities within the attack behavioral sequence. However, they typically overlook the specific threat information inherent in visual modalities, which preserves key threat details from inherently multimodal CTI reports. Inspired by the remarkable multimodal understanding capabilities of Multimodal Large Language Models (MLLMs), we explore their potential in enhancing multimodal attack graph construction. To be specific, we propose a novel framework, MM-AttacKG, which can effectively extract key information from threat images and integrate it into attack graph construction, thereby enhancing the comprehensiveness and accuracy of attack graphs. It first employs a threat image parsing module to extract critical threat information from images and generate textual descriptions using MLLMs. Subsequently, it builds an iterative question-answering pipeline tailored for image parsing to refine the understanding of threat images. Finally, it achieves content-level integration between attack graphs and image-based answers through MLLMs, completing threat information enhancement. We construct a new multimodal dataset, AG-LLM-mm, and conduct extensive experiments to evaluate the effectiveness of MM-AttacKG. The results demonstrate that MM-AttacKG can accurately identify key information in threat images and significantly improve the quality of multimodal attack graph construction, effectively addressing the shortcomings of existing methods in utilizing image-based threat information.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"338 ","pages":"Article 115483"},"PeriodicalIF":7.6,"publicationDate":"2026-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175184","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

Federated vision transformer with adaptive focal loss for medical image classification 用于医学图像分类的自适应焦损联邦视觉变压器

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-04 DOI: 10.1016/j.knosys.2026.115474

Xinyuan Zhao , Yihang Wu , Ahmad Chaddad , Tareef Daqqaq , Reem Kateb

While deep learning models like Vision Transformer (ViT) have achieved significant advances, they typically require large datasets. With data privacy regulations, access to many original datasets is restricted, especially medical images. Federated learning (FL) addresses this challenge by enabling global model aggregation without data exchange. However, the heterogeneity of the data and the class imbalance that exist in local clients pose challenges for the generalization of the model. This study proposes a FL framework leveraging a dynamic adaptive focal loss (DAFL) and a client-aware aggregation strategy for local training. Specifically, we design a dynamic class imbalance coefficient that adjusts based on each client’s sample distribution and class data distribution, ensuring minority classes receive sufficient attention and preventing sparse data from being ignored. To address client heterogeneity, a weighted aggregation strategy is adopted, which adapts to data size and characteristics to better capture inter-client variations. The classification results on three public datasets (ISIC, Ocular Disease and RSNA-ICH) show that the proposed framework outperforms DenseNet121, ResNet50, ViT-S/16, ViT-L/32, FedCLIP, Swin Transformer, CoAtNet, and MixNet in most cases, with accuracy improvements ranging from 0.98% to 41.69%. Ablation studies on the imbalanced ISIC dataset validate the effectiveness of the proposed loss function and aggregation strategy compared to traditional loss functions and other FL approaches. The codes can be found at: https://github.com/AIPMLab/ViT-FLDAF.

虽然像Vision Transformer （ViT）这样的深度学习模型已经取得了重大进展，但它们通常需要大型数据集。由于数据隐私法规的限制，对许多原始数据集的访问受到限制，尤其是医学图像。联邦学习（FL）通过支持无需数据交换的全局模型聚合来解决这一挑战。然而，本地客户中存在的数据异质性和类别不平衡给模型的泛化带来了挑战。本研究提出了一个利用动态自适应焦点丢失（DAFL）和客户感知聚合策略进行局部训练的FL框架。具体而言，我们设计了一个动态类失衡系数，根据每个客户端的样本分布和类数据分布进行调整，保证少数类得到足够的关注，防止稀疏数据被忽略。为了解决客户端异构问题，采用了加权聚合策略，该策略可根据数据大小和特征进行调整，从而更好地捕获客户端之间的差异。在ISIC、Ocular Disease和RSNA-ICH三个公共数据集上的分类结果表明，在大多数情况下，所提出的框架优于DenseNet121、ResNet50、viti - s /16、viti - l /32、FedCLIP、Swin Transformer、CoAtNet和MixNet，准确率提高了0.98% ~ 41.69%。在不平衡ISIC数据集上的消融研究验证了所提出的损失函数和聚合策略与传统损失函数和其他FL方法相比的有效性。这些代码可以在https://github.com/AIPMLab/ViT-FLDAF上找到。

{"title":"Federated vision transformer with adaptive focal loss for medical image classification","authors":"Xinyuan Zhao , Yihang Wu , Ahmad Chaddad , Tareef Daqqaq , Reem Kateb","doi":"10.1016/j.knosys.2026.115474","DOIUrl":"10.1016/j.knosys.2026.115474","url":null,"abstract":"<div><div>While deep learning models like Vision Transformer (ViT) have achieved significant advances, they typically require large datasets. With data privacy regulations, access to many original datasets is restricted, especially medical images. Federated learning (FL) addresses this challenge by enabling global model aggregation without data exchange. However, the heterogeneity of the data and the class imbalance that exist in local clients pose challenges for the generalization of the model. This study proposes a FL framework leveraging a dynamic adaptive focal loss (DAFL) and a client-aware aggregation strategy for local training. Specifically, we design a dynamic class imbalance coefficient that adjusts based on each client’s sample distribution and class data distribution, ensuring minority classes receive sufficient attention and preventing sparse data from being ignored. To address client heterogeneity, a weighted aggregation strategy is adopted, which adapts to data size and characteristics to better capture inter-client variations. The classification results on three public datasets (ISIC, Ocular Disease and RSNA-ICH) show that the proposed framework outperforms DenseNet121, ResNet50, ViT-S/16, ViT-L/32, FedCLIP, Swin Transformer, CoAtNet, and MixNet in most cases, with accuracy improvements ranging from 0.98% to 41.69%. Ablation studies on the imbalanced ISIC dataset validate the effectiveness of the proposed loss function and aggregation strategy compared to traditional loss functions and other FL approaches. The codes can be found at: <span><span>https://github.com/AIPMLab/ViT-FLDAF</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"338 ","pages":"Article 115474"},"PeriodicalIF":7.6,"publicationDate":"2026-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175230","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

Multi-objective optimization approach with decomposition-based algorithm for selecting tagSNPs 基于分解算法的标签snp选择多目标优化方法

IF 7.6 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2026-02-04 DOI: 10.1016/j.knosys.2026.115471

María Victoria Díaz-Galián, Miguel A. Vega-Rodríguez, Sergio Santander-Jiménez

Nowadays multiple bioinformatics issues can be solved by using evolutionary computation due to its potential to address complex optimization problems. TagSNP selection lies within this class of challenging problems, since genotyping all the Single Nucleotide Polymorphisms (SNPs) for haplotype identification is economically costly and time-consuming. If a reduced number of tagSNPs is chosen instead, the classification of haplotypes will accordingly show a worsening. As a result, tagSNP selection can be considered as a multi-objective optimization problem, in which the aim is to optimize haplotype dissimilarity while minimizing the number of selected tagSNPs. We propose and detail an approach based on the Multi-Objective Evolutionary Algorithm based on Decomposition (MOEA/D) for accurately selecting tagSNPs attending to these two objectives. The proposed method includes novel problem-aware operators for the initialization, crossover, and mutation to boost optimization capabilities. The proposal is experimentally compared with six approaches from the literature on five real datasets, using in the evaluation three quality metrics and their corresponding statistical analyses. The attained results denote that our algorithm provides statistically-significant improvements over previous methods with competitive runtimes, thus highlighting the relevance of the proposed multi-objective approach.

目前，由于进化计算具有解决复杂优化问题的潜力，许多生物信息学问题可以通过进化计算来解决。标签snp选择属于这类具有挑战性的问题，因为对所有单核苷酸多态性（snp）进行基因分型以进行单倍型鉴定既经济又耗时。如果选择较少数量的标签snp，单倍型的分类将相应地恶化。因此，标签snp选择可以看作是一个多目标优化问题，其目的是优化单倍型不相似性，同时使选择的标签snp数量最小化。我们提出并详细介绍了一种基于基于分解的多目标进化算法（MOEA/D）的方法，用于准确选择符合这两个目标的标记snp。提出的方法包括新的问题感知算子，用于初始化、交叉和突变，以提高优化能力。在5个真实数据集上，将该方法与文献中的6种方法进行了实验比较，使用了3种质量度量及其相应的统计分析。所获得的结果表明，我们的算法在具有竞争运行时间的情况下提供了统计上显着的改进，从而突出了所提出的多目标方法的相关性。

{"title":"Multi-objective optimization approach with decomposition-based algorithm for selecting tagSNPs","authors":"María Victoria Díaz-Galián, Miguel A. Vega-Rodríguez, Sergio Santander-Jiménez","doi":"10.1016/j.knosys.2026.115471","DOIUrl":"10.1016/j.knosys.2026.115471","url":null,"abstract":"<div><div>Nowadays multiple bioinformatics issues can be solved by using evolutionary computation due to its potential to address complex optimization problems. TagSNP selection lies within this class of challenging problems, since genotyping all the Single Nucleotide Polymorphisms (SNPs) for haplotype identification is economically costly and time-consuming. If a reduced number of tagSNPs is chosen instead, the classification of haplotypes will accordingly show a worsening. As a result, tagSNP selection can be considered as a multi-objective optimization problem, in which the aim is to optimize haplotype dissimilarity while minimizing the number of selected tagSNPs. We propose and detail an approach based on the Multi-Objective Evolutionary Algorithm based on Decomposition (MOEA/D) for accurately selecting tagSNPs attending to these two objectives. The proposed method includes novel problem-aware operators for the initialization, crossover, and mutation to boost optimization capabilities. The proposal is experimentally compared with six approaches from the literature on five real datasets, using in the evaluation three quality metrics and their corresponding statistical analyses. The attained results denote that our algorithm provides statistically-significant improvements over previous methods with competitive runtimes, thus highlighting the relevance of the proposed multi-objective approach.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"338 ","pages":"Article 115471"},"PeriodicalIF":7.6,"publicationDate":"2026-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146135625","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