Knowledge-Based Systems最新文献_第4页

Accelerating one-shot neural architecture search via constructing a sparse search space 通过构建稀疏搜索空间加速单次神经架构搜索

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-23 DOI: 10.1016/j.knosys.2024.112620

Hongtao Huang , Xiaojun Chang , Lina Yao

Neural Architecture Search (NAS) has garnered significant attention for its ability to automatically design high-quality deep neural networks (DNNs) tailored to various hardware platforms. The major challenge for NAS is the time-consuming network estimation process required to select optimal networks from a large pool of candidates. Rather than training each candidate from scratch, recent one-shot NAS methods accelerate the estimation process by only training a supernet and sampling sub-networks from it, inheriting partial network architectures and weights. Despite significant acceleration, the supernet training with a large search space (i.e., the number of candidate sub-networks) still requires thousands of GPU hours to support high-quality sub-network sampling. In this work, we propose SparseNAS, an approach for one-shot NAS acceleration by reducing the redundancy of the search space. We observe that many sub-networks in the space are underperforming, with significant performance disparity to high-performance sub-networks. Crucially, this disparity can be observed early in the beginning of the supernet training. Therefore, we train an early predictor to learn this disparity and filter out high-quality networks in advance. Then, the supernet training will be conducted in this space sub-space. Compared to the state-of-the-art one-shot NAS, our SparseNAS reports a

3.1 \times

training speedup with comparable network performance on the ImageNet dataset. Compared to the state-of-the-art acceleration method, SparseNAS reports a maximum of 1.5% higher Top-1 accuracy and 28% training cost reduction with a

7 \times

bigger search space. Extensive experiment results demonstrated that SparseNAS achieves better trade-offs between efficiency and performance than state-of-the-art one-shot NAS.

神经架构搜索（NAS）能够自动设计适合各种硬件平台的高质量深度神经网络（DNN），因而备受关注。NAS 面临的主要挑战是从大量候选网络中选择最佳网络所需的耗时网络估计过程。最近的单次 NAS 方法不是从头开始训练每个候选网络，而是只训练一个超级网络并从中抽取子网络，继承部分网络架构和权重，从而加速了估计过程。尽管加速效果显著，但具有较大搜索空间（即候选子网络数量）的超级网络训练仍需要数千个 GPU 小时才能支持高质量的子网络采样。在这项工作中，我们提出了 SparseNAS，这是一种通过减少搜索空间冗余来实现单次 NAS 加速的方法。我们观察到，该空间中的许多子网络性能不佳，与高性能子网络的性能差距很大。最重要的是，这种差距在超级网络训练初期就能观察到。因此，我们训练了一个早期预测器来学习这种差异，并提前筛选出高质量的网络。然后，超网训练将在这个空间子空间中进行。与最先进的单次NAS相比，我们的SparseNAS在ImageNet数据集上的训练速度提高了3.1倍，网络性能相当。与最先进的加速方法相比，SparseNAS 的 Top-1 准确率最高提高了 1.5%，训练成本降低了 28%，搜索空间扩大了 7 倍。广泛的实验结果表明，与最先进的单次NAS相比，SparseNAS在效率和性能之间实现了更好的权衡。

{"title":"Accelerating one-shot neural architecture search via constructing a sparse search space","authors":"Hongtao Huang , Xiaojun Chang , Lina Yao","doi":"10.1016/j.knosys.2024.112620","DOIUrl":"10.1016/j.knosys.2024.112620","url":null,"abstract":"<div><div>Neural Architecture Search (NAS) has garnered significant attention for its ability to automatically design high-quality deep neural networks (DNNs) tailored to various hardware platforms. The major challenge for NAS is the time-consuming network estimation process required to select optimal networks from a large pool of candidates. Rather than training each candidate from scratch, recent one-shot NAS methods accelerate the estimation process by only training a supernet and sampling sub-networks from it, inheriting partial network architectures and weights. Despite significant acceleration, the supernet training with a large search space (i.e., the number of candidate sub-networks) still requires thousands of GPU hours to support high-quality sub-network sampling. In this work, we propose SparseNAS, an approach for one-shot NAS acceleration by reducing the redundancy of the search space. We observe that many sub-networks in the space are underperforming, with significant performance disparity to high-performance sub-networks. Crucially, this disparity can be observed early in the beginning of the supernet training. Therefore, we train an <em>early predictor</em> to learn this disparity and filter out high-quality networks in advance. Then, the supernet training will be conducted in this space sub-space. Compared to the state-of-the-art one-shot NAS, our SparseNAS reports a <span><math><mrow><mn>3</mn><mo>.</mo><mn>1</mn><mo>×</mo></mrow></math></span> training speedup with comparable network performance on the ImageNet dataset. Compared to the state-of-the-art acceleration method, SparseNAS reports a maximum of 1.5% higher Top-1 accuracy and 28% training cost reduction with a <span><math><mrow><mn>7</mn><mo>×</mo></mrow></math></span> bigger search space. Extensive experiment results demonstrated that SparseNAS achieves better trade-offs between efficiency and performance than state-of-the-art one-shot NAS.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112620"},"PeriodicalIF":7.2,"publicationDate":"2024-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142573332","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

PPDistiller: Weakly-supervised 3D point cloud semantic segmentation framework via point-to-pixel distillation PPDistiller：通过点到像素蒸馏实现弱监督三维点云语义分割框架

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-23 DOI: 10.1016/j.knosys.2024.112654

Yong Zhang , Zhaolong Wu , Rukai Lan , Yingjie Liang , Yifan Liu

Despite the significant growth in the availability of 3D light detection and ranging (LiDAR) point cloud data in recent years, annotation remains expensive and time-consuming. This has led to an increasing demand for weakly-supervised semantic segmentation (WSSS) methods in applications such as autonomous driving, mapping, and robotics. Existing approaches typically rely solely on LiDAR point cloud data for WSSS, which often results in lower segmentation accuracy due to the sparsity of point clouds. To address these challenges, we propose a novel architecture, PPDistiller, which employs multiple teacher networks from different modalities. Compared to other WSSS and multimodal approaches, PPDistiller achieves superior segmentation accuracy with fewer annotations. This is facilitated through the novel Mean Multi-Teacher Framework (MMT), which incorporates multiple modalities and teachers. To address the issue of lacking 2D labels, we propose the Distance-CAM Self-Training (DCAM-ST) module, which utilizes sparse 3D weak annotations to produce accurate 2D pixel-level annotations. To enable adaptive fusion of 2D and 3D data, we introduce the Attention Point to Pixel Fusion (APPF) module, facilitating bidirectional transfer of cross-modal knowledge. Additionally, to fully leverage the spatial semantic information in point cloud, we propose the Pyramid Semantic-context Neighbor Aggregation (PSNA) module, aiming to exploit spatial and semantic correlations to improve performance. Extensive experimentation on SemanticKITTI, ScribbleKITTI and nuScenes datasets demonstrates the superiority of our proposed method. Compared to state-of-the-art fusion and weakly-supervised methods, PPDistiller achieves the highest mean Intersection over Union (mIoU) scores under both fully-supervised and weakly-supervised settings.

尽管近年来三维光探测与测距（LiDAR）点云数据的可用性大幅提高，但标注仍然昂贵且耗时。这导致自动驾驶、制图和机器人等应用领域对弱监督语义分割（WSSS）方法的需求不断增加。现有的 WSSS 方法通常仅依赖于激光雷达点云数据，由于点云的稀疏性，通常会导致较低的分割精度。为了应对这些挑战，我们提出了一种新型架构 PPDistiller，它采用了来自不同模态的多个教师网络。与其他 WSSS 和多模态方法相比，PPDistiller 以更少的注释实现了更高的分割精度。这得益于新颖的平均多教师框架（MMT），该框架结合了多种模式和教师。为了解决缺乏二维标签的问题，我们提出了距离-CAM 自我训练（DCAM-ST）模块，该模块利用稀疏的三维弱注释来生成准确的二维像素级注释。为了实现二维和三维数据的自适应融合，我们引入了注意力点到像素融合（APPF）模块，以促进跨模态知识的双向传输。此外，为了充分利用点云中的空间语义信息，我们提出了金字塔语义上下文邻居聚合（PSNA）模块，旨在利用空间和语义相关性来提高性能。在 SemanticKITTI、ScribbleKITTI 和 nuScenes 数据集上进行的大量实验证明了我们提出的方法的优越性。与最先进的融合方法和弱监督方法相比，PPDistiller 在完全监督和弱监督设置下都获得了最高的平均交叉联合（mIoU）分数。

{"title":"PPDistiller: Weakly-supervised 3D point cloud semantic segmentation framework via point-to-pixel distillation","authors":"Yong Zhang , Zhaolong Wu , Rukai Lan , Yingjie Liang , Yifan Liu","doi":"10.1016/j.knosys.2024.112654","DOIUrl":"10.1016/j.knosys.2024.112654","url":null,"abstract":"<div><div>Despite the significant growth in the availability of 3D light detection and ranging (LiDAR) point cloud data in recent years, annotation remains expensive and time-consuming. This has led to an increasing demand for weakly-supervised semantic segmentation (WSSS) methods in applications such as autonomous driving, mapping, and robotics. Existing approaches typically rely solely on LiDAR point cloud data for WSSS, which often results in lower segmentation accuracy due to the sparsity of point clouds. To address these challenges, we propose a novel architecture, PPDistiller, which employs multiple teacher networks from different modalities. Compared to other WSSS and multimodal approaches, PPDistiller achieves superior segmentation accuracy with fewer annotations. This is facilitated through the novel Mean Multi-Teacher Framework (MMT), which incorporates multiple modalities and teachers. To address the issue of lacking 2D labels, we propose the Distance-CAM Self-Training (DCAM-ST) module, which utilizes sparse 3D weak annotations to produce accurate 2D pixel-level annotations. To enable adaptive fusion of 2D and 3D data, we introduce the Attention Point to Pixel Fusion (APPF) module, facilitating bidirectional transfer of cross-modal knowledge. Additionally, to fully leverage the spatial semantic information in point cloud, we propose the Pyramid Semantic-context Neighbor Aggregation (PSNA) module, aiming to exploit spatial and semantic correlations to improve performance. Extensive experimentation on SemanticKITTI, ScribbleKITTI and nuScenes datasets demonstrates the superiority of our proposed method. Compared to state-of-the-art fusion and weakly-supervised methods, PPDistiller achieves the highest mean Intersection over Union (mIoU) scores under both fully-supervised and weakly-supervised settings.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112654"},"PeriodicalIF":7.2,"publicationDate":"2024-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142553807","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-scale feature reconstruction network for industrial anomaly detection 用于工业异常检测的多尺度特征重构网络

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-23 DOI: 10.1016/j.knosys.2024.112650

Ehtesham Iqbal , Samee Ullah Khan , Sajid Javed , Brain Moyo , Yahya Zweiri , Yusra Abdulrahman

Unsupervised anomaly detection techniques, which operate without prior knowledge of anomalies, have garnered significant attention in industrial inspection due to their adaptability and generalization. Therefore, knowledge-based computer vision techniques have been broadly applied to identify unusual image patterns. However, real-time industrial applications present challenges such as limited anomalous samples, inadequate defect knowledge, and complex background textures. These factors lead to difficulties in accurately identifying defect regions, and conventional auto-encoder networks often struggle to overcome these issues.To address these limitations, we propose a multi-scale feature reconstruction (MSFR) network specifically designed for domain shift scenarios. Our approach employs a pyramidal vision transformer network (PVTN) to reconstruct multi-scale feature maps, capturing discriminative features at various scales. Additionally, a pre-trained module extracts multi-level features at the same scale, and a dedicated feature matching module enhances accuracy by improving the alignment probability between features. The MSFR strategy surpasses conventional auto-encoders by filtering pixel-level information at multiple depths. Empirical evaluations were conducted using benchmark datasets such as MVTec AD and AeBAD-S. Furthermore, an extensive ablation study demonstrates the effectiveness and viability of the proposed MSFR approach for industrial anomaly detection tasks. The experimental results show that the proposed model significantly outperforms recent approaches, making it highly suitable for real-world industrial applications, particularly in manufacturing.

无监督异常检测技术无需事先了解异常情况，因其适应性和通用性，在工业检测领域备受关注。因此，基于知识的计算机视觉技术已被广泛应用于识别异常图像模式。然而，实时工业应用面临着异常样本有限、缺陷知识不足和背景纹理复杂等挑战。为了解决这些局限性，我们提出了一种专为域偏移场景设计的多尺度特征重构（MSFR）网络。我们的方法采用金字塔视觉变换器网络（PVTN）来重建多尺度特征图，捕捉不同尺度的鉴别特征。此外，预训练模块可提取同一尺度的多级特征，而专用的特征匹配模块则可通过提高特征间的配准概率来提高准确性。MSFR 策略通过过滤多个深度的像素级信息，超越了传统的自动编码器。使用 MVTec AD 和 AeBAD-S 等基准数据集进行了经验评估。此外，一项广泛的消融研究证明了所提出的 MSFR 方法在工业异常检测任务中的有效性和可行性。实验结果表明，所提出的模型明显优于最近的方法，因此非常适合实际工业应用，尤其是制造业。

{"title":"Multi-scale feature reconstruction network for industrial anomaly detection","authors":"Ehtesham Iqbal , Samee Ullah Khan , Sajid Javed , Brain Moyo , Yahya Zweiri , Yusra Abdulrahman","doi":"10.1016/j.knosys.2024.112650","DOIUrl":"10.1016/j.knosys.2024.112650","url":null,"abstract":"<div><div>Unsupervised anomaly detection techniques, which operate without prior knowledge of anomalies, have garnered significant attention in industrial inspection due to their adaptability and generalization. Therefore, knowledge-based computer vision techniques have been broadly applied to identify unusual image patterns. However, real-time industrial applications present challenges such as limited anomalous samples, inadequate defect knowledge, and complex background textures. These factors lead to difficulties in accurately identifying defect regions, and conventional auto-encoder networks often struggle to overcome these issues.To address these limitations, we propose a multi-scale feature reconstruction (MSFR) network specifically designed for domain shift scenarios. Our approach employs a pyramidal vision transformer network (PVTN) to reconstruct multi-scale feature maps, capturing discriminative features at various scales. Additionally, a pre-trained module extracts multi-level features at the same scale, and a dedicated feature matching module enhances accuracy by improving the alignment probability between features. The MSFR strategy surpasses conventional auto-encoders by filtering pixel-level information at multiple depths. Empirical evaluations were conducted using benchmark datasets such as MVTec AD and AeBAD-S. Furthermore, an extensive ablation study demonstrates the effectiveness and viability of the proposed MSFR approach for industrial anomaly detection tasks. The experimental results show that the proposed model significantly outperforms recent approaches, making it highly suitable for real-world industrial applications, particularly in manufacturing.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112650"},"PeriodicalIF":7.2,"publicationDate":"2024-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142573327","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Automatic medical report generation combining contrastive learning and feature difference 结合对比学习和特征差异自动生成医疗报告

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-23 DOI: 10.1016/j.knosys.2024.112630

Chongwen Lyu , Chengjian Qiu , Kai Han , Saisai Li , Victor S. Sheng , Huan Rong , Yuqing Song , Yi Liu , Zhe Liu

The automatic medical report generation is a challenging task because it requires accurate capture and description of abnormal regions, especially for those discrepancies between patient and normal. In most cases, normal region descriptions dominate the entire medical report, and existing methods may fail to focus on abnormal regions due to data bias. Medical reports can be automatically generated by combining contrastive learning with feature difference in order to capture and describe abnormal regions effectively. By capturing discrepancy attributes between the input image and normal images, this method can provide more accurate diagnostic reports and better represent the visual features of abnormal regions. Specifically, we propose the feature difference approach to make the model focus more on abnormal regions, and on the other hand, we propose the combination of contrastive learning for enhancing the visual representation of feature difference based on the feature difference approach, thus improving the performance of the model. Experimental results on the IU-Xray and MIMIC-CXR datasets demonstrate the effectiveness of our approach.

自动生成医疗报告是一项具有挑战性的任务，因为它需要准确捕捉和描述异常区域，尤其是患者与正常人之间的差异。在大多数情况下，正常区域的描述在整个医疗报告中占主导地位，而现有的方法可能会由于数据偏差而无法关注异常区域。通过将对比学习与特征差异相结合，可以自动生成医疗报告，从而有效捕捉和描述异常区域。通过捕捉输入图像与正常图像之间的差异属性，该方法可以提供更准确的诊断报告，并更好地表现异常区域的视觉特征。具体来说，我们一方面提出了特征差异方法，使模型更加关注异常区域；另一方面，我们在特征差异方法的基础上，提出了结合对比学习来增强特征差异的视觉表示，从而提高模型的性能。在 IU-Xray 和 MIMIC-CXR 数据集上的实验结果证明了我们方法的有效性。

{"title":"Automatic medical report generation combining contrastive learning and feature difference","authors":"Chongwen Lyu , Chengjian Qiu , Kai Han , Saisai Li , Victor S. Sheng , Huan Rong , Yuqing Song , Yi Liu , Zhe Liu","doi":"10.1016/j.knosys.2024.112630","DOIUrl":"10.1016/j.knosys.2024.112630","url":null,"abstract":"<div><div>The automatic medical report generation is a challenging task because it requires accurate capture and description of abnormal regions, especially for those discrepancies between patient and normal. In most cases, normal region descriptions dominate the entire medical report, and existing methods may fail to focus on abnormal regions due to data bias. Medical reports can be automatically generated by combining contrastive learning with feature difference in order to capture and describe abnormal regions effectively. By capturing discrepancy attributes between the input image and normal images, this method can provide more accurate diagnostic reports and better represent the visual features of abnormal regions. Specifically, we propose the feature difference approach to make the model focus more on abnormal regions, and on the other hand, we propose the combination of contrastive learning for enhancing the visual representation of feature difference based on the feature difference approach, thus improving the performance of the model. Experimental results on the IU-Xray and MIMIC-CXR datasets demonstrate the effectiveness of our approach.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112630"},"PeriodicalIF":7.2,"publicationDate":"2024-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142537612","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-interest sequential recommendation with contrastive learning and temporal analysis 利用对比学习和时间分析进行多兴趣顺序推荐

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-23 DOI: 10.1016/j.knosys.2024.112657

Xiaowen Ma , Qiang Zhou , Yongjun Li

Sequential recommendation systems aim to forecast the subsequent item of interest to users by analyzing their historical behaviors. While existing approaches, which employ attention mechanisms, have significantly advanced by capturing users’ multiple interests, they encounter two primary challenges. Firstly, they often fail to effectively capture the transient shifts in users’ interests across a sequence of items and neglect the interdependencies among these items, leading to a misalignment between the identified and actual interests. Secondly, conventional multi-interest models struggle to ensure that the identified interests are distinct, which results in overly similar interests that may not adequately satisfy user requirements. To address these issues, we propose a novel multi-interest recommendation method, which models the temporal features and user’s preference features from the user level. In order to capture short-term variations in interest, we introduce a time period module to encode the behavioral intervals between items and capture the periodicity of users clicking on similar items by extracting temporal information. In addition, we integrate similar types of items into the interest subgraph through preference feature extraction to capture users’ short-term changes in relevance term interests, and incorporate contrastive learning to enhance the differences between the captured interests. Extensive experiments conducted on two datasets Amazon Books and Taobao show that the model outperforms current state-of-the-art methods.

顺序推荐系统旨在通过分析用户的历史行为，预测用户感兴趣的后续项目。现有的方法采用了注意力机制，通过捕捉用户的多重兴趣取得了显著进步，但也遇到了两个主要挑战。首先，它们往往无法有效捕捉用户兴趣在一系列项目中的瞬时变化，也忽视了这些项目之间的相互依存关系，从而导致所识别的兴趣与实际兴趣不一致。其次，传统的多兴趣模型难以确保所识别的兴趣是不同的，从而导致兴趣过于相似，无法充分满足用户需求。为了解决这些问题，我们提出了一种新颖的多兴趣推荐方法，该方法从用户层面对时间特征和用户偏好特征进行建模。为了捕捉兴趣的短期变化，我们引入了一个时间段模块来编码项目之间的行为间隔，并通过提取时间信息来捕捉用户点击相似项目的周期性。此外，我们还通过偏好特征提取将相似类型的项目整合到兴趣子图中，以捕捉用户对相关词兴趣的短期变化，并结合对比学习来增强捕捉到的兴趣之间的差异。在亚马逊图书和淘宝网两个数据集上进行的大量实验表明，该模型优于目前最先进的方法。

{"title":"Multi-interest sequential recommendation with contrastive learning and temporal analysis","authors":"Xiaowen Ma , Qiang Zhou , Yongjun Li","doi":"10.1016/j.knosys.2024.112657","DOIUrl":"10.1016/j.knosys.2024.112657","url":null,"abstract":"<div><div>Sequential recommendation systems aim to forecast the subsequent item of interest to users by analyzing their historical behaviors. While existing approaches, which employ attention mechanisms, have significantly advanced by capturing users’ multiple interests, they encounter two primary challenges. Firstly, they often fail to effectively capture the transient shifts in users’ interests across a sequence of items and neglect the interdependencies among these items, leading to a misalignment between the identified and actual interests. Secondly, conventional multi-interest models struggle to ensure that the identified interests are distinct, which results in overly similar interests that may not adequately satisfy user requirements. To address these issues, we propose a novel multi-interest recommendation method, which models the temporal features and user’s preference features from the user level. In order to capture short-term variations in interest, we introduce a time period module to encode the behavioral intervals between items and capture the periodicity of users clicking on similar items by extracting temporal information. In addition, we integrate similar types of items into the interest subgraph through preference feature extraction to capture users’ short-term changes in relevance term interests, and incorporate contrastive learning to enhance the differences between the captured interests. Extensive experiments conducted on two datasets Amazon Books and Taobao show that the model outperforms current state-of-the-art methods.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112657"},"PeriodicalIF":7.2,"publicationDate":"2024-10-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142553802","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

Neural network algorithm with transfer learning and dropout for using a UAV to search the lost target in motion 使用无人飞行器搜索运动中丢失目标的带有迁移学习和丢弃功能的神经网络算法

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-22 DOI: 10.1016/j.knosys.2024.112632

Yiying Zhang , Guanzhong Zhou , Chao Huang , Hailong Huang

When performing the rescue mission, how to find a lost target in motion as soon as possible is a very valuable and interesting research topic. To find the optimal flight path of the UAV used to search the lost target in motion under different prior knowledge and interference intensity, this paper proposes a novel optimization technique called neural network algorithm with transfer learning and dropout (TLDNNA), which is a variant of neural network algorithm (NNA) inspired by artificial neural networks. To improve the global search ability and convergence performance of NNA, the multiple transfer mechanism, the generalized mean transfer position, and the random dropout operator driven by the principles of transfer learning and dropout are introduced to TLDNNA. To investigate the performance of TLDNNA, TLDNNA and 12 powerful population-based optimization algorithms are applied to the problem of using a UAV to search the lost target in motion in the designed 9 scenarios. Experimental results show that TLDNNA is remarkably better than the other 12 algorithms in most of the scenarios in terms of solution quality, stability, and convergence performance. In addition, TLDNNA can improve the overall search performance by 18.14% and overall computational efficiency by 3.41 times compared to NNA. The source code of TLDNNA can be obtained by https://github.com/jsuzyy/TLDNNA.

在执行救援任务时，如何尽快找到运动中的失联目标是一个非常有价值和有趣的研究课题。为了在不同的先验知识和干扰强度下找到用于搜索运动中的迷失目标的无人机的最优飞行路径，本文提出了一种新颖的优化技术，称为带转移学习和丢弃的神经网络算法（TLDNNA），它是受人工神经网络启发的神经网络算法（NNA）的一个变种。为了提高 NNA 的全局搜索能力和收敛性能，在 TLDNNA 中引入了多重转移机制、广义平均转移位置以及由转移学习和丢弃原理驱动的随机丢弃算子。为了研究 TLDNNA 的性能，将 TLDNNA 和 12 种强大的基于种群的优化算法应用于设计的 9 个场景中使用无人机搜索运动中丢失目标的问题。实验结果表明，在大多数情况下，TLDNNA 在解的质量、稳定性和收敛性方面都明显优于其他 12 种算法。此外，与 NNA 相比，TLDNNA 的整体搜索性能提高了 18.14%，整体计算效率提高了 3.41 倍。TLDNNA 的源代码可通过 https://github.com/jsuzyy/TLDNNA 获取。

{"title":"Neural network algorithm with transfer learning and dropout for using a UAV to search the lost target in motion","authors":"Yiying Zhang , Guanzhong Zhou , Chao Huang , Hailong Huang","doi":"10.1016/j.knosys.2024.112632","DOIUrl":"10.1016/j.knosys.2024.112632","url":null,"abstract":"<div><div>When performing the rescue mission, how to find a lost target in motion as soon as possible is a very valuable and interesting research topic. To find the optimal flight path of the UAV used to search the lost target in motion under different prior knowledge and interference intensity, this paper proposes a novel optimization technique called neural network algorithm with transfer learning and dropout (TLDNNA), which is a variant of neural network algorithm (NNA) inspired by artificial neural networks. To improve the global search ability and convergence performance of NNA, the multiple transfer mechanism, the generalized mean transfer position, and the random dropout operator driven by the principles of transfer learning and dropout are introduced to TLDNNA. To investigate the performance of TLDNNA, TLDNNA and 12 powerful population-based optimization algorithms are applied to the problem of using a UAV to search the lost target in motion in the designed 9 scenarios. Experimental results show that TLDNNA is remarkably better than the other 12 algorithms in most of the scenarios in terms of solution quality, stability, and convergence performance. In addition, TLDNNA can improve the overall search performance by 18.14% and overall computational efficiency by 3.41 times compared to NNA. The source code of TLDNNA can be obtained by <span><span>https://github.com/jsuzyy/TLDNNA</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112632"},"PeriodicalIF":7.2,"publicationDate":"2024-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142553806","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 multi-aware graph convolutional network for driver drowsiness detection 用于检测驾驶员瞌睡的多感知图卷积网络

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-21 DOI: 10.1016/j.knosys.2024.112643

Liang Lin , Song Wang , Jucheng Yang , Feng Wei

Driver drowsiness is a leading cause of traffic accidents. Utilizing deep neural networks, facial feature-based methods have achieved promising results in drowsiness detection. However, these methods suffer from two limitations. Firstly, they only focus on features from one or two facial regions, thus overlooking the relationships between features across different facial regions. Secondly, these methods struggle to account for individual driver variability, a common phenomenon where drivers may display dissimilar signs of drowsiness. These limitations lead to inaccurate drowsiness detection. To address these issues, in this paper we propose a multi-aware graph convolutional network (MAGCN). At the heart of MAGCN are two feature extractors: the class- and attention-aware extractor (CAAE), and the composite temporal-aware extractor (CTAE). The CAAE explores interdependencies within global and local facial features, while the CTAE leverages temporal information to capture dynamic changes in features. Moreover, a task-oriented graph convolutional network is designed to refine the drowsiness feature space for precise detection. Experiment results show that the proposed MAGCN exhibits competitive detection performance, when compared with state-of-the-art drowsiness detection approaches on two public datasets. In summary, the proposed model not only learns and analyzes correlations between features from various facial regions, but also tackles individual driver variability.

驾驶员瞌睡是交通事故的主要原因。利用深度神经网络，基于面部特征的方法在嗜睡检测方面取得了可喜的成果。然而，这些方法存在两个局限性。首先，它们只关注一个或两个面部区域的特征，从而忽略了不同面部区域特征之间的关系。其次，这些方法难以考虑驾驶员的个体差异，而驾驶员可能会表现出不同的嗜睡迹象，这是一种常见现象。这些局限性导致嗜睡检测不准确。为了解决这些问题，我们在本文中提出了一种多感知图卷积网络（MAGCN）。MAGCN 的核心是两个特征提取器：类别和注意力感知提取器（CAAE）和复合时间感知提取器（CTAE）。CAAE 探索全局和局部面部特征之间的相互依存关系，而 CTAE 则利用时间信息来捕捉特征的动态变化。此外，还设计了面向任务的图卷积网络，以完善嗜睡特征空间，从而实现精确检测。实验结果表明，在两个公共数据集上，与最先进的嗜睡检测方法相比，所提出的 MAGCN 具有极具竞争力的检测性能。总之，所提出的模型不仅能学习和分析不同面部区域特征之间的相关性，还能解决驾驶员个体差异问题。

{"title":"A multi-aware graph convolutional network for driver drowsiness detection","authors":"Liang Lin , Song Wang , Jucheng Yang , Feng Wei","doi":"10.1016/j.knosys.2024.112643","DOIUrl":"10.1016/j.knosys.2024.112643","url":null,"abstract":"<div><div>Driver drowsiness is a leading cause of traffic accidents. Utilizing deep neural networks, facial feature-based methods have achieved promising results in drowsiness detection. However, these methods suffer from two limitations. Firstly, they only focus on features from one or two facial regions, thus overlooking the relationships between features across different facial regions. Secondly, these methods struggle to account for individual driver variability, a common phenomenon where drivers may display dissimilar signs of drowsiness. These limitations lead to inaccurate drowsiness detection. To address these issues, in this paper we propose a multi-aware graph convolutional network (MAGCN). At the heart of MAGCN are two feature extractors: the class- and attention-aware extractor (CAAE), and the composite temporal-aware extractor (CTAE). The CAAE explores interdependencies within global and local facial features, while the CTAE leverages temporal information to capture dynamic changes in features. Moreover, a task-oriented graph convolutional network is designed to refine the drowsiness feature space for precise detection. Experiment results show that the proposed MAGCN exhibits competitive detection performance, when compared with state-of-the-art drowsiness detection approaches on two public datasets. In summary, the proposed model not only learns and analyzes correlations between features from various facial regions, but also tackles individual driver variability.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112643"},"PeriodicalIF":7.2,"publicationDate":"2024-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142537611","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

Cross-modal recipe retrieval based on unified text encoder with fine-grained contrastive learning 基于统一文本编码器和细粒度对比学习的跨模态食谱检索

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-21 DOI: 10.1016/j.knosys.2024.112641

Bolin Zhang , Haruya Kyutoku , Keisuke Doman , Takahiro Komamizu , Ichiro Ide , Jiangbo Qian

Cross-modal recipe retrieval is vital for transforming visual food cues into actionable cooking guidance, making culinary creativity more accessible. Existing methods separately encode the recipe Title, Ingredient, and Instruction using different text encoders, then aggregate them to obtain recipe feature, and finally match it with encoded image feature in a joint embedding space. These methods perform well but require significant computational cost. In addition, they only consider matching the entire recipe and the image but ignore the fine-grained correspondence between recipe components and the image, resulting in insufficient cross-modal interaction. To this end, we propose Unified Text Encoder with Fine-grained Contrastive Learning (UTE-FCL) to achieve a simple but efficient model. Specifically, in each recipe, UTE-FCL first concatenates each of the Ingredient and Instruction texts composed of multiple sentences as a single text. Then, it connects these two concatenated texts with the original single-phrase Title to obtain the concatenated recipe. Finally, it encodes these three concatenated texts and the original Title by a Transformer-based Unified Text Encoder (UTE). This proposed structure greatly reduces the memory usage and improves the feature encoding efficiency. Further, we propose fine-grained contrastive learning objectives to capture the correspondence between recipe components and the image at Title, Ingredient, and Instruction levels by measuring the mutual information. Extensive experiments demonstrate the effectiveness of UTE-FCL compared to existing methods.

跨模态食谱检索对于将视觉食物线索转化为可操作的烹饪指导至关重要，它使烹饪创意更易实现。现有的方法使用不同的文本编码器分别对菜谱标题、配料和说明进行编码，然后将其汇总以获得菜谱特征，最后在联合嵌入空间中将其与编码的图像特征进行匹配。这些方法性能良好，但需要大量计算成本。此外，它们只考虑了整个菜谱与图像的匹配，却忽略了菜谱组成部分与图像之间的细粒度对应关系，导致跨模态交互不足。为此，我们提出了具有细粒度对比学习功能的统一文本编码器（UTE-FCL），以实现简单而高效的模型。具体来说，在每个食谱中，UTE-FCL 首先将由多个句子组成的 "成分 "文本和 "说明 "文本串联为一个文本。然后，它将这两个串联文本与原始的单句标题连接起来，得到串联食谱。最后，它通过基于变换器的统一文本编码器（UTE）对这三个串联文本和原始标题进行编码。这种结构大大减少了内存使用量，提高了特征编码效率。此外，我们还提出了细粒度对比学习目标，通过测量互信息来捕捉食谱成分与图像在标题、成分和说明层面的对应关系。大量实验证明，与现有方法相比，UTE-FCL 非常有效。

{"title":"Cross-modal recipe retrieval based on unified text encoder with fine-grained contrastive learning","authors":"Bolin Zhang , Haruya Kyutoku , Keisuke Doman , Takahiro Komamizu , Ichiro Ide , Jiangbo Qian","doi":"10.1016/j.knosys.2024.112641","DOIUrl":"10.1016/j.knosys.2024.112641","url":null,"abstract":"<div><div>Cross-modal recipe retrieval is vital for transforming visual food cues into actionable cooking guidance, making culinary creativity more accessible. Existing methods separately encode the recipe Title, Ingredient, and Instruction using different text encoders, then aggregate them to obtain recipe feature, and finally match it with encoded image feature in a joint embedding space. These methods perform well but require significant computational cost. In addition, they only consider matching the entire recipe and the image but ignore the fine-grained correspondence between recipe components and the image, resulting in insufficient cross-modal interaction. To this end, we propose <strong>U</strong>nified <strong>T</strong>ext <strong>E</strong>ncoder with <strong>F</strong>ine-grained <strong>C</strong>ontrastive <strong>L</strong>earning (UTE-FCL) to achieve a simple but efficient model. Specifically, in each recipe, UTE-FCL first concatenates each of the Ingredient and Instruction texts composed of multiple sentences as a single text. Then, it connects these two concatenated texts with the original single-phrase Title to obtain the concatenated recipe. Finally, it encodes these three concatenated texts and the original Title by a Transformer-based Unified Text Encoder (UTE). This proposed structure greatly reduces the memory usage and improves the feature encoding efficiency. Further, we propose fine-grained contrastive learning objectives to capture the correspondence between recipe components and the image at Title, Ingredient, and Instruction levels by measuring the mutual information. Extensive experiments demonstrate the effectiveness of UTE-FCL compared to existing methods.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112641"},"PeriodicalIF":7.2,"publicationDate":"2024-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142537616","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Unification of probabilistic graph model and deep reinforcement learning (UPGMDRL) for multi-intersection traffic signal control 统一概率图模型和深度强化学习（UPGMDRL）用于多交叉口交通信号控制

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-20 DOI: 10.1016/j.knosys.2024.112663

Ali Reza Sattarzadeh, Pubudu N. Pathirana

Traffic signals play a pivotal role in modern life by preventing collisions, regulating traffic flow, and ensuring a predictable and efficient transportation system. Adaptive traffic light signal control (ATSC) is a promising paradigm for mitigating traffic congestion in Intelligent Transportation Systems (ITS). Among various AI-based approaches, Deep Reinforcement Learning (DRL) has gained widespread application, demonstrating superior performance. This paper aims to develop a latent space reinforcement learning method for intelligent traffic control, with a focus on making explainable decisions. According to the latent model and hidden Markov mixed model, this method integrated both to develop an ATSC framework for traffic networks with multiple intersections. Given the challenges posed by high-dimensional data and a limited understanding of the task, traditional decision-making methods often struggle with understanding the environment. This paper aims to provide semantic information and an enhanced understanding of the environment by offering interpretable states. The latent model is employed to extract task-relevant information from underlying representations within a framework that unifies representation learning and DRL. The experimental results demonstrate how our approach effectively and efficiently balances traffic flow, leading to improved traffic management.

交通信号在现代生活中发挥着举足轻重的作用，它可以防止碰撞、调节交通流量，并确保交通系统的可预测性和高效性。自适应交通信号灯控制（ATSC）是智能交通系统（ITS）中缓解交通拥堵的一种前景广阔的范例。在各种基于人工智能的方法中，深度强化学习（DRL）已获得广泛应用，并展现出卓越的性能。本文旨在为智能交通控制开发一种潜空间强化学习方法，重点在于做出可解释的决策。根据潜模型和隐马尔可夫混合模型，该方法将二者结合起来，为具有多个交叉口的交通网络开发了一个 ATSC 框架。鉴于高维数据带来的挑战和对任务的有限理解，传统决策方法往往在理解环境方面举步维艰。本文旨在通过提供可解释的状态来提供语义信息并增强对环境的理解。在表征学习和 DRL 相结合的框架内，采用潜模型从底层表征中提取与任务相关的信息。实验结果表明了我们的方法如何有效、高效地平衡交通流量，从而改善交通管理。

{"title":"Unification of probabilistic graph model and deep reinforcement learning (UPGMDRL) for multi-intersection traffic signal control","authors":"Ali Reza Sattarzadeh, Pubudu N. Pathirana","doi":"10.1016/j.knosys.2024.112663","DOIUrl":"10.1016/j.knosys.2024.112663","url":null,"abstract":"<div><div>Traffic signals play a pivotal role in modern life by preventing collisions, regulating traffic flow, and ensuring a predictable and efficient transportation system. Adaptive traffic light signal control (ATSC) is a promising paradigm for mitigating traffic congestion in Intelligent Transportation Systems (ITS). Among various AI-based approaches, Deep Reinforcement Learning (DRL) has gained widespread application, demonstrating superior performance. This paper aims to develop a latent space reinforcement learning method for intelligent traffic control, with a focus on making explainable decisions. According to the latent model and hidden Markov mixed model, this method integrated both to develop an ATSC framework for traffic networks with multiple intersections. Given the challenges posed by high-dimensional data and a limited understanding of the task, traditional decision-making methods often struggle with understanding the environment. This paper aims to provide semantic information and an enhanced understanding of the environment by offering interpretable states. The latent model is employed to extract task-relevant information from underlying representations within a framework that unifies representation learning and DRL. The experimental results demonstrate how our approach effectively and efficiently balances traffic flow, leading to improved traffic management.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112663"},"PeriodicalIF":7.2,"publicationDate":"2024-10-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142529399","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

HeteroKGRep: Heterogeneous Knowledge Graph based Drug Repositioning HeteroKGRep：基于异构知识图谱的药物重新定位

IF 7.2 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Knowledge-Based Systems

Pub Date : 2024-10-19 DOI: 10.1016/j.knosys.2024.112638

Ribot Fleury T. Ceskoutsé , Alain Bertrand Bomgni , David R. Gnimpieba Zanfack , Diing D.M. Agany , Bouetou Bouetou Thomas , Etienne Gnimpieba Zohim

The process of developing new drugs is both time-consuming and costly, often taking over a decade and billions of dollars to obtain regulatory approval. Additionally, the complexity of patent protection for novel compounds presents challenges for pharmaceutical innovation. Drug repositioning offers an alternative strategy to uncover new therapeutic uses for existing medicines. Previous repositioning models have been limited by their reliance on homogeneous data sources, failing to leverage the rich information available in heterogeneous biomedical knowledge graphs. We propose HeteroKGRep, a novel drug repositioning model that utilizes heterogeneous graphs to address these limitations. HeteroKGRep is a multi-step framework that first generates a similarity graph from hierarchical concept relations. It then applies SMOTE over-sampling to address class imbalance before generating node sequences using a heterogeneous graph neural network. Drug and disease embeddings are extracted from the network and used for prediction. We evaluated HeteroKGRep on a graph containing biomedical concepts and relations from ontologies, pathways and literature. It achieved state-of-the-art performance with 99% accuracy, 95% AUC ROC and 94% average precision on predicting repurposing opportunities. Compared to existing homogeneous approaches, HeteroKGRep leverages diverse knowledge sources to enrich representation learning. Based on heterogeneous graphs, HeteroKGRep can discover new drug-disease associations, leveraging de novo drug development. This work establishes a promising new paradigm for knowledge-guided drug repositioning using multimodal biomedical data.

开发新药的过程既耗时又耗资，通常需要十多年的时间和数十亿美元才能获得监管部门的批准。此外，新型化合物专利保护的复杂性也给制药创新带来了挑战。药物重新定位为现有药物发掘新的治疗用途提供了另一种策略。以往的重新定位模型由于依赖同质数据源而受到限制，无法充分利用异质生物医学知识图谱中的丰富信息。我们提出的 HeteroKGRep 是一种新型药物重新定位模型，它利用异构图来解决这些局限性。HeteroKGRep 是一个多步骤框架，首先从分层概念关系中生成一个相似性图。然后，在使用异构图神经网络生成节点序列之前，它应用 SMOTE 过度采样来解决类不平衡问题。从网络中提取药物和疾病嵌入，并用于预测。我们对 HeteroKGRep 进行了评估，该图包含本体、路径和文献中的生物医学概念和关系。它在预测再利用机会方面达到了最先进的性能，准确率为 99%，AUC ROC 为 95%，平均精度为 94%。与现有的同质方法相比，HeteroKGRep 利用不同的知识源来丰富表征学习。基于异构图谱，HeteroKGRep 可以发现新的药物-疾病关联，充分利用新药开发。这项工作为利用多模态生物医学数据进行知识引导的药物重新定位建立了一个前景广阔的新范例。

{"title":"HeteroKGRep: Heterogeneous Knowledge Graph based Drug Repositioning","authors":"Ribot Fleury T. Ceskoutsé , Alain Bertrand Bomgni , David R. Gnimpieba Zanfack , Diing D.M. Agany , Bouetou Bouetou Thomas , Etienne Gnimpieba Zohim","doi":"10.1016/j.knosys.2024.112638","DOIUrl":"10.1016/j.knosys.2024.112638","url":null,"abstract":"<div><div>The process of developing new drugs is both time-consuming and costly, often taking over a decade and billions of dollars to obtain regulatory approval. Additionally, the complexity of patent protection for novel compounds presents challenges for pharmaceutical innovation. Drug repositioning offers an alternative strategy to uncover new therapeutic uses for existing medicines. Previous repositioning models have been limited by their reliance on homogeneous data sources, failing to leverage the rich information available in heterogeneous biomedical knowledge graphs. We propose HeteroKGRep, a novel drug repositioning model that utilizes heterogeneous graphs to address these limitations. HeteroKGRep is a multi-step framework that first generates a similarity graph from hierarchical concept relations. It then applies SMOTE over-sampling to address class imbalance before generating node sequences using a heterogeneous graph neural network. Drug and disease embeddings are extracted from the network and used for prediction. We evaluated HeteroKGRep on a graph containing biomedical concepts and relations from ontologies, pathways and literature. It achieved state-of-the-art performance with 99% accuracy, 95% AUC ROC and 94% average precision on predicting repurposing opportunities. Compared to existing homogeneous approaches, HeteroKGRep leverages diverse knowledge sources to enrich representation learning. Based on heterogeneous graphs, HeteroKGRep can discover new drug-disease associations, leveraging de novo drug development. This work establishes a promising new paradigm for knowledge-guided drug repositioning using multimodal biomedical data.</div></div>","PeriodicalId":49939,"journal":{"name":"Knowledge-Based Systems","volume":"305 ","pages":"Article 112638"},"PeriodicalIF":7.2,"publicationDate":"2024-10-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142529400","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