检索结果-内蒙古大学图书馆

One Person, One Model-Learning Compound Router for Sequential Recommendation

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Liu, Zhiding Cheng, Mingyue Li, Zhi Liu, Qi Chen, Enhong Anhui Province Key Laboratory of Big Data Analysis and Application University of Science and Technology of China Hefei China State Key Laboratory of Cognitive Intelligence Hefei China Shenzhen International Graduate School Tsinghua University Shenzhen China

Deep learning has brought significant breakthroughs in sequential recommendation (SR) for capturing dynamic user interests. A series of recent research revealed that models with more parameters usually achieve optimal performance for SR tasks, inevitably resulting in great challenges for deploying them in real systems. Following the simple assumption that light networks might already suffice for certain users, in this work, we propose CANet, a conceptually simple yet very scalable framework for assigning adaptive network architecture in an input-dependent manner to reduce unnecessary computation. The core idea of CANet is to route the input user behaviors with a light-weighted router module. Specifically, we first construct the routing space with various submodels parameterized in terms of multiple model dimensions such as the number of layers, hidden size and embedding size. To avoid extra storage overhead of the routing space, we employ a weight-slicing schema to maintain all the submodels in exactly one network. Furthermore, we leverage several solutions to solve the discrete optimization issues caused by the router module. Thanks to them, CANet could adaptively adjust its network architecture for each input in an end-to-end manner, in which the user preference can be effectively captured. To evaluate our work, we conduct extensive experiments on benchmark datasets. Experimental results show that CANet reduces computation by 55 ∼ 65% while preserving the accuracy of the original model. Our codes are available at https://***/icantnamemyself/CANet. Copyright © 2022, The Authors. All rights reserved.

关键词： Network architecture

Equivariant Graph Hierarchy-Based Neural Networks

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Han, Jiaqi Huang, Wenbing Xu, Tingyang Rong, Yu Department of Computer Science and Technology Tsinghua University China Gaoling School of Artificial Intelligence Renmin University of China China Beijing Key Laboratory of Big Data Management and Analysis Methods Beijing China Tencent AI Lab China

Equivariant Graph neural Networks (EGNs) are powerful in characterizing the dynamics of multi-body physical systems. Existing EGNs conduct flat message passing, which, yet, is unable to capture the spatial/dynamical hierarchy for complex systems particularly, limiting substructure discovery and global information fusion. In this paper, we propose Equivariant Hierarchy-based Graph Networks (EGHNs) which consist of the three key components: generalized Equivariant Matrix Message Passing (EMMP), E-Pool and E-UnPool. In particular, EMMP is able to improve the expressivity of conventional equivariant message passing, E-Pool assigns the quantities of the low-level nodes into high-level clusters, while E-UnPool leverages the high-level information to update the dynamics of the low-level nodes. As their names imply, both E-Pool and E-UnPool are guaranteed to be E(n)-equivariant to meet the physical symmetry. Considerable experimental evaluations verify the effectiveness of our EGHN on several applications including multi-object dynamics simulation, motion capture, and protein dynamics modeling. © 2022, CC BY.

关键词： Message passing

On efficient expanding training datasets of breast tumor ultrasound segmentation model

学校读者我要写书评

暂无评论

Computers in Biology and Medicine 2024年 183卷 109274-109274页

作者： Zhang, Caicai Mei, Mei Mei, Zhuolin Wu, Bin Chen, Shasha Lu, Minfeng Lu, Chenglang School of Modern Information Technology Zhejiang Polytechnic University of Mechanical and Electrical Engineering 528 Binwen Road Binjiang District Zhejiang Hangzhou310053 China Department of Ultrasound The Second Affiliated Hospital Zhejiang University School of Medicine 88 Jiefang Road Shangcheng District Zhejiang Hangzhou310009 China School of Computer and Big Data Science Jiujiang University 551 Qianjin East Road Jiangxi Jiujiang332005 China

Automatic segmentation of breast tumor ultrasound images can provide doctors with objective and efficient references for lesions and regions of interest. Both dataset optimization and model structure optimization are crucial for achieving optimal image segmentation performance, and it can be challenging to satisfy the clinical needs solely through model structure enhancements in the context of insufficient breast tumor ultrasound datasets for model training. While significant research has focused on enhancing the architecture of deep learning models to improve tumor segmentation performance, there is a relative paucity of work dedicated to dataset augmentation. Current data augmentation techniques, such as rotation and transformation, often yield insufficient improvements in model accuracy. The deep learning methods used for generating synthetic images, such as GANs is primarily applied to produce visually natural-looking images. Nevertheless, the accuracy of the labels for these generated images still requires manual verification, and the images exhibit a lack of diversity. Therefore, they are not suitable for the training datasets augmentation of image segmentation models. This study introduces a novel dataset augmentation approach that generates synthetic images by embedding tumor regions into normal images. We explore two synthetic methods: one using identical backgrounds and another with varying backgrounds. Through experimental validation, we demonstrate the efficiency of the synthetic datasets in enhancing the performance of image segmentation models. Notably, the synthetic method utilizing different backgrounds exhibits superior improvement compared to the identical background approach. Our findings contribute to medical image analysis, particularly in tumor segmentation, by providing a practical and effective dataset augmentation strategy that can significantly improve the accuracy and reliability of segmentation models. © 2024 Elsevier Ltd

关键词： Ultrasonic imaging

Regional adversarial training for better robust generalization

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Song, Chuanbiao Fan, Yanbo Yang, Yichen Wu, Baoyuan Li, Yiming Li, Zhifeng He, Kun Huazhong University of Science and Technology Tencent The Chinese University of Hong Kong Shenzhen China Shenzhen Research Institute of Big Data Tsinghua University

Adversarial training (AT) has been demonstrated as one of the most promising defense methods against various adversarial attacks. To our knowledge, existing AT-based methods usually train with the locally most adversarial perturbed points and treat all the perturbed points equally, which may lead to considerably weaker adversarial robust generalization on test data. In this work, we introduce a new adversarial training framework that considers the diversity as well as characteristics of the perturbed points in the vicinity of benign samples. To realize the framework, we propose a Regional Adversarial Training (RAT) defense method that first utilizes the attack path generated by the typical iterative attack method of projected gradient descent (PGD), and constructs an adversarial region based on the attack path. Then, RAT samples diverse perturbed training points efficiently inside this region, and utilizes a distance-aware label smoothing mechanism to capture our intuition that perturbed points at different locations should have different impact on the model performance. Extensive experiments on several benchmark datasets show that RAT consistently makes significant improvement on standard adversarial training (SAT), and exhibits better robust generalization. Copyright © 2021, The Authors. All rights reserved.

关键词： Gradient methods

Towards on Blockchain data Privacy Protection with Cryptography and Software Architecture Approach 2nd

学校读者我要写书评

暂无评论

Towards on Blockchain Data Privacy Protection with Cryptogra...

2nd International Conference on Blockchain and Trustworthy Systems, Blocksys 2020

作者： Wang, Zexu Wen, Bin Luo, Ziqiang School of Information Science and Technology Hainan Normal University Haikou China Cloud Computing and Big Data Research Center Hainan Normal University Haikou China Key Laboratory of Data Science and Intelligence Education of Ministry of Education Hainan Normal University Haikou571158 China

ISBN: (纸本)9789811592126

The essence of blockchain is to solve trust problems and realize value transfer. The traditional centralized processing method adopts centralized transmission and storage of users’ data privacy, which improves the security and reliability of processing. The non central blockchain technology uses distributed ledger technology to realize the characteristics of disintermediation, data tampering, traceability, work audit, etc. encryption algorithm is used to encrypt the data, and consensus mechanism makes the data sharing of blockchain system more fair and stable. The current passive data privacy schemes are basically based on cryptography. On the premise of satisfying the constraint mechanism of blockchain, this paper studies the technical framework, encryption mechanism and empirical analysis of blockchain, and discusses the information privacy protection method of hiding the original big data as much as possible, so as to improve the system performance and protect the security and privacy of user data. At the same time, we study data privacy protection from the perspective of software architecture, and propose a data privacy protection scheme through algorithm decomposition multi center collaboration method, which provides guidance for big data sharing and transaction based on blockchain. © 2020, Springer Nature Singapore Pte Ltd.

关键词： Blockchain

Quantitative prediction and evaluation of geothermal resource areas in the southwest section of the Mid-Spine Belt of Beautiful China

学校读者我要写书评

暂无评论

International Journal of Digital Earth 2022年第1期15卷 748-769页

作者： Zhe Chen Ruichun Chang Wenbo Zhao Sijia Li Huadong Guo Keyan Xiao Lin Wu Dong Hou Lu Zou College of Mathematics and Physics Chengdu University of TechnologyChengduPeople’s Republic of China Digital Hu Line Research Institute Chengdu University of TechnologyChengduPeople’s Republic of China Geomathematics Key Laboratory of Sichuan Province ChengduPeople’s Republic of China Key Laboratory of Digital Earth Science Chinese Academy of SciencesAerospace Information Research InstituteBeijingPeople’s Republic of China University of Chinese Academy of Sciences BeijingPeople’s Republic of China International Research Centre of Big Data for Sustainable Development Goals (CBAS) BeijingPeople’s Republic of China MNR Key Laboratory of Metallogeny and Mineral Assessment Institute of Mineral ResourcesChinese Academy of Geological SciencesBeijingPeople’s Republic of China

The geothermal resources in the southwest section of the Mid-Spine Belt of Beautiful China are abundant,but the quantitative prediction and evaluation of geothermal resources are very difficult. Based on geographic information system (GIS) and remote sensing (RS) platforms,six impact factors,namely land surface temperature,fault density,Gutenberg–Liszt B value,formation combination entropy,distance to river and aeromagnetic anomaly were selected. Through the establishment of the certainty factor model (CF),weights of the information entropy certainty factor model (ICF) and weights of the evidence certainty factor model (ECF),the geothermal potential in the study area were predicted quantitatively. Based on the ECF results,the six main geothermal resource areas were delineated. The results show that (1) ECF had high prediction accuracy (success index is 0.00405%,area ratio is 0.867);(2) The geothermal resource areas obtained were Ganzi–Ya’an–Liangshan,Panzhihua–Liangshan,Dali–Chuxiong,Nujiang–Baoshan,Diqing–Dali,and Lijiang–Diqing. The results provide a basis for the effective development and utilization of geothermal resources in the southwest section of the mid-ridge belt.

关键词： Mid-Spine Belt of Beautiful China certainty factor geothermal resource area weights of evidence

A deep learning-based embedding framework for object detection and recognition in underwater marine organisms

学校读者我要写书评

暂无评论

Research Square

Research Square 2021年

作者： Zhu, Jinde Fujian Provincial Key Lab of Big Data Mining and Applications School of Computer Science and Mathematics Fujian University of Technology Fuzhou350118 China

The detection of marine organisms is an important part of the intelligent strategy in marine ranch, which requires an underwater robot to detect the marine organism quickly and accurately in the complex ocean environment. Based on the latest deep learning arithmetic, this paper put forward to find the marine organism in a picture or video to construct a real-time objective invention system for marine organisms. The neural network arithmetic： YOLOv4 was employed to extract the deep features of marine organisms, implementing the accurate detection and size detection of different fish can use arithmetic for evaluation in fisheries. Furthermore, improving the architecture of the backbone and the neck connection is called YOLOv4-embedding. As a result, compared with other object detection arithmetic, YOLOv4-embedding object detection arithmetic was better at detection accuracy-higher detection confidence and higher detection ratio than other one-stage object detection arithmetic, EfficientDet-D3 example. The consequence demonstrates that the suggested instrument could implement the rapid invention of different varieties in marine organisms. Compared to the YOLOv4, the mAP75 of the YOLOv4-embedding achieves an improvement of 2.92% for the marine organism dataset at a rapid rate of ~51 FPS on RTX 3090, 60.8% AP50 for the MS COCO dataset. © 2021, CC BY.

关键词： Fish

Hadamard Attention Recurrent Transformer: A Strong Baseline for Stereo Matching Transformer

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Chen, Ziyang Zhang, Yongjun Li, Wenting Wang, Bingshu Wu, Yabo Zhao, Yong Chen, C.L. Philip College of Computer Science The State Key Laboratory of Public Big Data Guizhou University Guiyang550025 China School of Information Engineering Guizhou University of Commerce Guiyang550021 China School of Software Northwestern Polytechnical University Xi’an710129 China Key Laboratory of Integrated Microsystems Shenzhen Graduate School Peking University Shenzhen518055 China School of Computer Science and Engineering South China University of Technology Guangzhou510641 China

In light of the advancements in transformer technology, extant research posits the construction of stereo transformers as a potential solution to the binocular stereo matching challenge. However, constrained by the low-rank bottleneck and quadratic complexity of attention mechanisms, stereo transformers still fail to demonstrate sufficient nonlinear expressiveness within a reasonable inference time. The lack of focus on key homonymous points renders the representations of such methods vulnerable to challenging conditions, including reflections and weak textures. Furthermore, a slow computing speed is not conducive to the application. To overcome these difficulties, we present the Hadamard Attention Recurrent Stereo Transformer (HART) that incorporates the following components: 1) For faster inference, we present a Hadamard product paradigm for the attention mechanism, achieving linear computational complexity. 2) We designed a Dense Attention Kernel (DAK) to amplify the differences between relevant and irrelevant feature responses. This allows HART to focus on important details. DAK also converts zero elements to non-zero elements to mitigate the reduced expressiveness caused by the low-rank bottleneck. 3) To compensate for the spatial and channel interaction missing in the Hadamard product, we propose MKOI to capture both global and local information through the interleaving of large and small kernel convolutions. Experimental results demonstrate the effectiveness of our HART. In reflective area, HART ranked 1st on the KITTI 2012 benchmark among all published methods at the time of submission. Code is available at https://***/ ZYangChen/HART. Copyright © 2025, The Authors. All rights reserved.

关键词： Hadamard transforms

A data Augmentation Method for Disease Diagnosis Based on Single Cell Omics data 2nd

学校读者我要写书评

暂无评论

A Data Augmentation Method for Disease Diagnosis Based on Si...

2nd International Conference on Applied Intelligence, ICAI 2024

作者： Zhuang, Han Zhang, Qinhu Ningbo Institute of Digital Twin Eastern Institute of Technology No. 568 Tongxin Road Zhuangshi Street Zhejiang Ningbo315201 China Big Data and Intelligent Computing Research Center Guangxi Academy of Science Nanning530007 China Institute for Regenerative Medicine Medical Innovation Center and State Key Laboratory of Cardiology Shanghai East Hospital School of Medicine Tongji University Shanghai200123 China

ISBN: (纸本)9789819619061

Single cell omics data has enabled bioinformatics methods to identify cell type, infer gene regulatory network and beyond. Besides these great advances, there is active research about how to predict patients’ diagnosis based on cell predicted disease status. In this study, we propose a computational method consisting of a resampling method and an AI model to perform disease diagnosis with single cell omics data. According to our analysis, our proposed method outperforms baseline methods on three public datasets and is effective for disease diagnosis. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： disease diagnosis resampling single cell omics