检索结果-内蒙古大学图书馆

Prior-guided dual-stage diabetic retinopathy grading model based on feature collaboration of lesion and vascular structure

引用

Expert Systems with Applications 2025年 285卷

作者： Guo, Xiaoxin Yang, Guangqi Xu, Chenfangqian Dong, Hongliang Hu, Xiaoying Che, Songtian Key Laboratory of Symbol Computation and Knowledge Engineering of Ministry of Education Jilin University 2699 Qianjin Street Changchun130012 China College of Computer Science and Technology Jilin University 2699 Qianjin Street Changchun130012 China Ophthalmology Department Bethune First Hospital of Jilin University Changchun Jilin130021 China Ophthalmology Department Bethune Second Hospital of Jilin University Changchun Jilin130041 China

In order to fully utilize lesion features and vascular structure and solve the problem of class imbalance, diabetes retinopathy (DR) grading is modeled as a dual-stage task, and the prior-guided dual-stage diabetes retinopathy grading model is proposed by designing an end-to-end hierarchical framework, consisting of XE-Net and MFC-Net. The former uses vascular structure from color fundus images as prior knowledge to guide the classification of proliferative diabetic retinopathy (PDR) and non-PDR;the latter further classifies non-PDR images using color fundus images with four types of segmented lesions. The core of MFC-Net is the feature collaboration block (FCB) for multi-classification tasks, which is proposed to facilitate effective collaboration between global and local features. In addition, a class-balanced (CB) loss weight is also introduced to address the common class imbalance problem in DR datasets. The experimental results on two publicly available datasets, DDR and IDRiD show that the proposed model has significant advantages. © 2025 Elsevier Ltd

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

Research and analysis of psychological data based on machine learning methods

引用

International Journal of Wireless and Mobile Computing 2022年第1期22卷 1-8页

作者： Chen, Guangshun Lv, Wei Ma, Junwei Liang, Yanchun Zhuhai Lab. of Key Lab. of Symbol Computation and Knowledge Engineering of Ministry of Education Zhuhai College of Jilin University Jilin Zhuhai China Key Laboratory of Symbol Computation and Knowledge Engineering of Ministry of Education College of Computer Science and Technology Jilin University Jilin Changchun China

The integration of psychology and computer science has become the mainstream contemporary research method on psychological data. Weibo, China's largest open platform for communication and information sharing between users, has many emotional contents hidden in its data. According to the current trend, the Weibo data are segmented by machine learning to obtain a psychological portrait of Weibo users. This design uses Long and Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNNs) to perform sentiment classification on Weibo data. The classification results are analysed using word frequency analysis and the Latent Dirichlet Allocation (LDA) model to obtain portraits of Weibo users' sentiment and an analysis of the results. The results are displayed in the form of word clouds. According to the clustering results of the word clouds, the main factors affecting different polar emotions can be analysed. © 2022 Inderscience Enterprises Ltd.. All rights reserved.

关键词： Statistics

来源：评论

学校读者我要写书评

暂无评论

Grid Partition and Agglomeration for Bidirectional Hierarchical Clustering 2nd

Grid Partition and Agglomeration for Bidirectional Hierarchi...

引用

2nd EAI International Conference on Security and Privacy in New Computing Environments, SPNCE 2019

作者： Wu, Lei Chen, Hechang Yu, Xiangchun Chao, Sun Yu, Zhezhou Dou, RuiTing College of Computer Science and Technology Jilin University Changchun China Key Laboratory of Symbolic Computation and Knowledge Engineering Ministry of Education Changchun China

ISBN: (纸本)9783030213725

Clustering is an important data processing tool, which can be used to reveal the distribution structure of unfamiliar domain data, or as preprocess methods to magnify data object to accelerate subsequent processing or simplify models. However, the distribution of many real-world data in feature space is very complex or uneven. Besides, the similarity/distance is not easy to be properly defined in feature space with different dimensional quantity. Therefore, many existing clustering algorithms are not stable in real datasets, and better performance of different datasets relies on artificial special design, such as scale normalization. In this paper, we propose a bidirectional hierarchical clustering (BHC) algorithm with two phases. In the first phase (Top-down), based on the probability density function of data in different dimensions, the feature space is divided into over-segmented grids to adapt to the complex distribution of data. In the second phase (Bottom-up), based on statistical information, a robust distance instead of geometrical distance is defined to agglomerate the grids into a dendrogram. Compared with the individual data points, grids created in the first phase can carry more statistical information, and the magnified processing objects can accelerate the clustering process. The second phase enhances the algorithm’s ability by the ability of recognize arbitrary shape data clusters. The effectiveness of BHC is compared with 20 popular or recent clustering algorithms on 8 artificial datasets and 6 real-world datasets. And the results show that our algorithm can achieve good results on most datasets. In particular, BHC surpasses all the comparison algorithms involved in the experiment on all real-world datasets. In addition, in order to test the efficiency of the algorithm, we design an experiment which can test the influence of dimension and data size on the operation time. © 2019, ICST Institute for Computer Sciences, Social Informatics and Telecommuni

关键词： Clustering algorithms

来源：评论

学校读者我要写书评

暂无评论

FlexPDA:A Flexible Programming Framework for Deep Learning Accelerators

引用

Journal of Computer Science & Technology 2022年第5期37卷 1200-1220页

作者： Lei Liu Xiu Ma Hua-xiao Liu Guang-li Li Lei Liu College of Computer Science and Technology Jilin UniversityChangchunChina Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education Jilin UniversityChangchunChina State Key Laboratory of Computer Architecture Institute of Computing TechnologyChinese Academy of SciencesBeijingChina University of Chinese Academy of Sciences BeijingChina

There are a wide variety of intelligence accelerators with promising performance and energy efficiency,deployed in a broad range of applications such as computer vision and speech ***,programming productivity hinders the deployment of deep learning *** low-level library invoked in the high-level deep learning framework which supports the end-to-end execution with a given model,is designed to reduce the programming burden on the intelligence ***,it is inflexible for developers to build a network model for every deep learning application,which probably brings unnecessary repetitive *** this paper,a flexible and efficient programming framework for deep learning accelerators,FlexPDA,is proposed,which provides more optimization opportunities than the low-level library and realizes quick transplantation of applications to intelligence accelerators for fast *** evaluate FlexPDA by using 10 representative operators selected from deep learning algorithms and an end-to-end *** experimental results validate the effectiveness of FlexPDA,which achieves an end-to-end performance improvement of 1.620x over the low-level library.

关键词： deep learning accelerator programming framework domain-specific language

来源：评论

学校读者我要写书评

暂无评论

Learning dynamic bayesian network with immune evolutionary algorithm

Learning dynamic bayesian network with immune evolutionary a...

引用

International Conference on Machine Learning and Cybernetics, ICMLC 2005

作者： Jia, Hai-Yang Liu, Da-You Yu, Peng Key Laboratory of Symbolic Computing and Knowledge Engineering Juin University Ministry of Education Changchun China

ISBN: (纸本)078039092X

Dynamic Bayesian Networks (DBNs) are directed graphical models of stochastic processes, How to learn the structure of DBNs from data is a hot problem of research. In this paper the author presents an Immune evolutionary algorithm for learning the network structure of DBNs. The results of contrast experiment prove that the constringency rate is more rapid than EGA-DBN algorithms. © 2005 IEEE.

关键词： Learning systems

来源：评论

学校读者我要写书评

暂无评论

Crossed-Time Delay Neural Network for Speaker Recognition 1

引用

27th International Conference on MultiMedia Modeling, MMM 2021

作者： Chen, Liang Liang, Yanchun Shi, Xiaoshu Zhou, You Wu, Chunguo Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education College of Computer Science and Technology Jilin University Changchun130012 China Zhuhai Laboratory of Key Laboratory of Symbol Computation and Knowledge Engineering of Ministry of Education School of Computer Zhuhai College of Jilin University Zhuhai519041 China

ISBN: (数字)9783030678326

ISBN: (纸本)9783030678319

Time Delay Neural Network (TDNN) is a well-performing structure for deep neural network-based speaker recognition systems. In this paper we introduce a novel structure, named Crossed-Time Delay Neural Network (CTDNN) to enhance the performance of current TDNN for speaker recognition. Inspired by the multi-filters setting of convolution layers from convolution neural networks, we set multiple time delay units with different context size at the bottom layer and construct a multilayer parallel network. The proposed CTDNN gives significant improvements over original TDNN on both speaker verification and identification tasks. It outperforms in VoxCeleb1 dataset in verification experiment with a 2.6% absolute Equal Error Rate improvement. In few shots condition, CTDNN reaches 90.4% identification accuracy, which doubles the identification accuracy of original TDNN. We also compare the proposed CTDNN with another new variant of TDNN, Factorized-TDNN, which shows that our model has a 36% absolute identification accuracy improvement under few shots condition. Moreover, the proposed CTDNN can handle training with a larger batch more efficiently and hence, utilize calculation resources economically. © 2021, Springer Nature Switzerland AG.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

CBR and micro-architecture antipatterns based software design improvement

引用

IFIP TC12 1st International Conference on Artificial Intelligence Applications and Innovations, AIAI 2004

作者： Feng, Tie Zhang, Jiachen Wang, Hongyuan Wang, Xian Institute of Computer Science and Technology Jilin University Changchun 130012 China Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education China

ISBN: (纸本)1402081502

This paper presents a Case Based Reasoning (CBR) approach to identifying micro-architecture anti-patterns and replacing them with "good" patterns in order to improve the design of software system. The resulting system design benefits from better flexibility for adapting to future requirement change and expansion. In this approach, both problematic inflexible structures and their corresponding refactoring designs are formally defined and organized in a case base. The identification of anti-patterns is carried out through similarity measurement on class diagrams, sequence diagrams, OO quality metric and semantic constraints. A supporting system CBDIT was developed to aid this approach. © 2004 Springer Science + Business Media, Inc.

关键词： Case based reasoning

来源：评论

学校读者我要写书评

暂无评论

Applying Kumaraswamy distribution on stick-breaking process: a Dirichlet neural topic model approach

引用

Neural Computing and Applications 2024年第22期36卷 13731-13744页

作者： Ouyang, Jihong Wang, Teng Cao, Jingyue Wang, Yiming College of Computer Science and Technology Jilin University Qianjin Street Jilin Changchun130015 China Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education Jilin University Qianjin Street Jilin Changchun130015 China

In recent years, neural topic modeling has increasingly raised extensive attention due to its capacity on generating coherent topics and flexible deep neural structures. However, the widely used Dirichlet distribution in shallow topic models is difficult to reparameterize. Therefore, most existing neural topic models assume the Gaussian as the prior of topic proportions for reparameterization. Gaussian distribution does not have the sparsity like Dirichlet distribution, which limits the model’s topic extraction ability. To address this issue, we propose a novel neural topic model approximating the Dirichlet prior with the reparameterizable Kumaraswamy distribution, namely Kumaraswamy Neural Topic Model (KNTM). Specifically, we adopted the stick-breaking process for posterior inference with the Kumaraswamy distribution as the base distribution. Besides, to capture the dependencies among topics, we propose a Kumaraswamy Recurrent Neural Topic Model (KRNTM) based on the recurrent stick-breaking construction to ensure that the model can still generate coherent topical words in high-dimensional topic space. We examined our method on five prevalent benchmark datasets over six Dirichlet-approximating neural topic models, among which KNTM has the lowest perplexity and KRNTM performance best on topic coherence and topic uniqueness. Qualitative analysis of the top topical words verifies that our proposed models can extract more semantically coherent topics compared with state-of-the-art models, further demonstrating our method’s effectiveness. This work contributes to the broader application of VAEs with Dirichlet priors. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Benchmarking

来源：评论

学校读者我要写书评

暂无评论

A new SAT Encoding Scheme for Exactly-one Constraints 5

A new SAT Encoding Scheme for Exactly-one Constraints

引用

5th Annual International Conference on Network and Information Systems for Computers, ICNISC 2019

作者： Cai, Jiatong Su, Yating Yang, Xixi Min, Jionghao Lai, Yong College of Software Jilin University Changchun China Key Laboratory of Symbolic Computation and Knowledge Engineering Ministry of Education Jilin University Changchun China

Exactly-one constraints have comprehensive applications for the fields of artificial intelligence and operations research. For many encoded SAT problems generated by the existing encoding schemes of exactly-one constraints, the state-of-the-art knowledge compilers cannot complete compilation. In this paper, we propose a new encoding scheme of exactly-one constraints. We introduce two-dimensional auxiliary variables (represented as a matrix) to denote the constraint that exactly one of some variables can be assigned as true. The clauses generated by our scheme is significantly less than those generated by three other existing encoding schemes. The experimental results on the exact cover problems show that the encoded CNF formulas generated by our scheme requires less compilation time, compared with the other three coding schemes. © Published under licence by IOP Publishing Ltd.

关键词： Operations research

来源：评论

学校读者我要写书评

暂无评论

Model for human dynamics based on habit

引用

Chinese Science Bulletin 2010年第24期55卷 2744-2749页

作者： JIAO Yu LIU YanHeng WANG Jian WANG Jing College of Computer Science and Technology Jilin University Changchun 130012 China Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education Jilin University Changchun 130012 China Archives of Jilin University Changchun 130012 China

The quantitative understanding of human behavior is a central question of modern science. Because of the complexity of human behavior, it is almost impossible to seek regularities in human dynamics. It is assumed that human actions are randomly distributed in time in current models for human dynamics. While the characteristics of human behavior combined with the queue model are considered as model for human dynamics based on habit to explain bursts and heavy tails in human dynamics more exactly. Normal distribution is used to simulate intervals of succession of events, and random parameters are set as unexpected events disturbing habit behaviors. Moreover, duration of events are proposed to imitate continual attention to some events in human behaviors.

关键词：人类行为基于模型动力学模型学习突发事件时间间隔现代科学随机分布

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：