检索结果-内蒙古大学图书馆

2025 IEEE International Conference on Acoustics, Speech, and Signal processing, ICASSP 2025

作者： Ti, Zheng Hao, Xuze Chen, Renhai College of Intelligence and Computing Tianjin University Tianjin China School of Computer Science Shanghai Key Laboratory of Intelligent Information Processing Fudan University Shanghai China

ISBN: (纸本)9798350368741

Continual Semantic Segmentation (CSS) aims to continuously learn new classes while mitigating catastrophic forgetting. Existing CSS methods primarily address this challenge through knowledge distillation. While they focus on maintaining stability for old classes, this emphasis often restricts plasticity in learning new ones. In this work, we propose two plug-and-play methods to better balance stability and plasticity. First, for feature map distillation, we replace pixel-level constraints with statistical information extraction within a window, improving the learning of new classes while preserving structural integrity to overcome forgetting. Second, we use relative ranks of predictions to replace exact probabilistic values to loosen the constraints in the logit distribution of old classes. Extensive experiments demonstrate that our methods outperform state-of-the-art methods when integrating into several baselines across various CSS scenarios. © 2025 IEEE.

关键词： Continual Learning Knowledge Distillation Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

Deep Reinforcement Learning based intelligent Routing for Multi-service Traffic SDN 5

Deep Reinforcement Learning based Intelligent Routing for Mu...

引用

5th International Academic Exchange Conference on Science and Technology Innovation, IAECST 2023

作者： Wang, Lingjiao Li, Jiachen Xiang Tan University Key Laboratory of Intelligent Computing and Information Processing Ministry of Education China Xiang Tan University School of Automation and Electronic Information China

ISBN: (纸本)9798350357738

The development of network has imposed higher requirements on QoS (Quality of Service) for routing, and the emergence of traffic classification techniques, software defined networking (SDN), and programmable network devices has made it possible to quickly identify user needs and route for service traffic requirements. Deep reinforcement learning routing algorithms for existing QoS compliant routing algorithms perform poorly in terms of generalizability. A deep graph reinforcement learning intelligent routing algorithm DDGL applied to SDN scenarios is proposed, which uses GNN to extract the link feature information of SDN data plane for deep reinforcement learning routing decision making, and comprehensively takes into account the performance metrics such as latency, packet loss, throughput, and bandwidth, to improve the algorithm generalization ability and enhance the SDN routing performance. © 2023 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

ArtDiff: Artwork Generation via Conditional Diffusion Models 23

ArtDiff: Artwork Generation via Conditional Diffusion Models

引用

15th International Conference on Digital Image processing, ICDIP 2023

作者： Liu, Wenhui Liu, Bin Lin, Xiao Li, Bo He, Zhifen Wang, Kang School of Mathematics and Information Science Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang Hangkong University Jiangxi Nanchang China Department of Supercomputing Management Jiangxi Science and Technology Infrastructure Center Jiangxi Nanchang China

ISBN: (纸本)9798400708237

Artwork Generation is an important research area of computer vision. Recently, kinds of generative models have achieved great success in natural image generation. However, artwork generation has rarely been studied due to the unfixed structure of artworks. Combined with prevailing diffusion models, we propose a simple yet effective framework, named as ArtDiff, for artwork generation. Given a name of an artist, we can generate diverse and novel artworks which reflecting the style of the artist. To the best of our knowledge, ArtDiff is the first diffusion model that generate artworks with the guidance of artist's name. Experimental results demonstrate that ArtDiff is able to recognize the artist's preference and generate artworks with reasonable structures and fine-grained details. © 2023 ACM.

关键词： Diffusion

来源：评论

学校读者我要写书评

暂无评论

Remote Sensing Images Semantic Segmentation Method Based on Improved Nested UNet

Remote Sensing Images Semantic Segmentation Method Based on ...

引用

2022 International Conference on Geographic information and Remote Sensing Technology, GIRST 2022

作者： Li, Zhongyu Liu, Yang Kuang, Yin Wang, Huajun Liu, Cheng College of Geophysics Chengdu University of Technology Chengdu610059 China College of Computer Science Chengdu Normal University Chengdu611130 China Key Laboratory of interior Layout optimization and Security Institutions of Higher Education of Sichuan Province Sichuan Chengdu611130 China Key Laboratory of Pattern Recognition and Intelligent Information Processing of Sichuan Chengdu University Chengdu610106 China Artificial Intelligence Key Laboratory of Sichuan Province Zigong643000 China

ISBN: (纸本)9781510662186

With the development of remote sensing technology, remote sensing images of buildings are of great significance in urban planning, disaster response, and other directions. When we use a neural network containing batch normalization layers for semantic segmentation, the neural network is sensitive to batch size and has low segmentation accuracy for occluded and dense buildings. This paper proposes a method for building segmentation in remote sensing images based on Nested UNet (UNet++) deep neural network. First, the UNet++ network is used to extract features, and the Group Normalization (GN) method is used instead of Batch Normalization (BN) to alleviate the model's sensitivity to batch size. Then the weighted combination of Cross-Entropy Loss (CELoss) and DiceLoss is used as the loss function to improve the feature extraction ability of the neural network for unbalanced buildings. Finally, experiments are carried out on the WHUBuilding dataset. The experimental results show that the improved model (UNet++-GN) improves Mean Intersection over Union (MIoU) and Mean Pixel Accuracy (Macc) by 12.16% and 2.92%, respectively, compared with the original model (UNet++-BN). © 2023 SPIE.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

Accelerating the HS-type Richardson Iteration Method with Anderson Mixing

引用

Acta Mathematica Sinica,English Series 2022年第11期38卷 2069-2089页

作者： Zhi Zhi LI Huai ZHANG Le OU-YANG Key Laboratory of Computational Geodynamics University of Chinese Academy of SciencesBeijing 100049P.R.China Guangdong Key Laboratory of Intelligent Information Processing Shenzhen Key Laboratory of Media Securityand Guangdong Laboratory of Artificial Intelligence and Digital Economy(SZ)College of Electronics and Information EngineeringShenzhen UniversityShenzhen 518060P.R.China

The Accelerated Hermitian/skew-Hermitian type Richardson(AHSR)iteration methods are presented for solving non-Hermitian positive definite linear systems with three schemes,by using Anderson *** upper bounds of spectral radii of iteration matrices are studied,and then the convergence theories of the AHSR iteration methods are ***,the optimal iteration parameters are provided,which can be computed *** addition,the application to the model convection-diffusion equation is depicted and numerical experiments are conducted to exhibit the effectiveness and confirm the theoretical analysis of the AHSR iteration methods.

关键词： Anderson mixing Hermitian/skew-Hermitian splitting the Richardson iteration convergence analyses optimal parameters the model convection-diffusion equation

来源：评论

学校读者我要写书评

暂无评论

An Enhanced Pythagorean Fuzzy Distance Method with Application in the Selection of Renewable Energy Source via MCDA

An Enhanced Pythagorean Fuzzy Distance Method with Applicati...

引用

2024 International Conference on New Trends in Computational Intelligence, NTCI 2024

作者： Terna, Anum Manasseh Ejegwa, Paul Augustine Feng, Yuming Joseph Sarwuan Tarka University Department of Mathematics Makurdi Nigeria Chongqing Three Gorges University Key Laboratory of Intelligent Information Processing and Control Chongqing Wanzhou404100 China

ISBN: (纸本)9798331517021

Many complex decision-making processes have been prosecuted with the aid of Pythagorean fuzzy distance methods. This is because the theory of Pythagorean fuzzy sets has the ability to handle the vagueness and imprecision in decision-making. Various Pythagorean fuzzy distance methods have been constructed and applied to multiple criteria decision-making. Recently, Thakur et al. developed a Pythagorean fuzzy distance method based on Euler's number by considering an incomplete number of the parameters of Pythagorean fuzzy sets. This exclusion could render the result of the distance method unreliable in real-life applications. In this work, we augment the Thakur et al:'s Pythagorean fuzzy distance method by the inclusion of the complete parameters of Pythagorean fuzzy sets for an effective interpretation. The properties of the enhanced Pythagorean fuzzy distance method are described and the method satisfies the conditions of distance function. In addition, the new method is used to discuss the selection of renewable energy source based on multiple criteria decision-making approach. Furthermore, the enhanced method is compared with other existing Pythagorean fuzzy distance methods via multiple criteria decision-making approach to show the superiority of the augmented method. © 2024 IEEE.

关键词： multi-criteria decision analysis multiple criteria decision-making Pythagorean fuzzy distance method Pythagorean fuzzy set renewable energy source

来源：评论

学校读者我要写书评

暂无评论

A topic relevance-aware click model for web search

引用

Journal of intelligent and Fuzzy Systems 2024年第4期46卷 8961-8974页

作者： Jianping, Liu Yingfei, Wang Jian, Wang Meng, Wang Xintao, Chu College of Computer Science and Engineering North Minzu University Yinchuan China The Key Laboratory of Images Graphics Intelligent Processing of State Ethnic Affairs Commission North Minzu University Yinchuan China Agricultural Information Institute Chinese Academy of Agricultural Sciences Beijing China

To better understand users’ behavior patterns in web search, numerous click models are proposed to extract the implicit interaction feedback. Most existing click models are heavily based on the implicit information to model user behaviors, ignoring the impact of explicit information between queries and documents in search sessions. In this paper, we fully consider the topic relevance between queries and documents in search sessions and propose a novel topic relevance-aware click model (TRA-CM) for web search. TRA-CM consists of a relevance estimator and an examination predictor. The relevance estimator consists of a topic relevance predictor and a click context encoder. In the topic relevance predictor, we utilize the pre-trained BERT model to model the content information of queries and documents in search sessions. Meanwhile, we use transformer to encode users’ click behaviors in the click context encoder. We further apply a two-stage fusion strategy to obtain the final relevance scores. The examination predictor estimates the examination probability of each document. We further utilize learnable filters to attenuate log noise and obtain purer input features in both relevance estimator and examination predictor, and investigate different combination functions to integrate relevance scores and examination probabilities into click prediction. Extensive experiment results on two real-world session datasets prove that TRA-CM outperforms existing click models in both click prediction and relevance estimation tasks. © 2024 – IOS Press. All rights reserved.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

A novel torus-based optical networks-on-chip exploiting double path communication algorithm 2

A novel torus-based optical networks-on-chip exploiting doub...

引用

2nd International Conference on Electronic information Engineering, Big Data, and Computer Technology, EIBDCT 2023

作者： Dai, Li Su, Ye Xie, Yiyuan School of Electronic and Information Engineering Southwest University Chongqing China Chongqing Key Laboratory of Nonlinear Circuits and Intelligent Information Processing Chongqing400715 China

ISBN: (纸本)9781510664968

Optical networks-on-chip (ONoCs) using optical media to interconnect cores, have the advantages of high bandwidth, low power consumption and low latency. At present, mesh, torus and fat-tree are the most widely used topologies in ONoCs. Torus structure has the advantages of higher connectivity due to its annular structure. However, the traditional torus structure still has the problem of high end-to-end (ETE) delay. This paper proposes a new double-path ONoCs based on torus topology, named DPtorus. The DPtorus consists of two layers: one layer is a torus network divided into unit clusters, the other layer is an all-pass optical router. The dual-layer network is connected with each other by optical vias, so that there are two communication pathes between the source node and the target node in the network can be selected. The One is inter-cluster communication and the other is non-cluster communication. According to the comparison of the number of hop of the two communication paths, the communication path with fewer hops between the nodes is preferably selected to reduce the communication cost of the data packet. The network simulation results show that the DPtorus reduces latency and improves throughput compared to the torus network, resulting in better performance and enhanced reliability of network communications for the DPtorus. © 2023 SPIE.

关键词： Topology

来源：评论

学校读者我要写书评

暂无评论

A video archive management method based on blockchain technology 11

A video archive management method based on blockchain techno...

引用

Optoelectronic Imaging and Multimedia Technology XI 2024

作者： Wu, Rongteng College of Computer and Big Data Minjiang University Fuzhou350108 China Fujian Provincial Key Laboratory of Information Processing and Intelligent Control Minjiang University Fuzhou350108 China

ISBN: (纸本)9781510682061

Video archives play a crucial role in various industries, from media and entertainment to security and surveillance. Managing these archives efficiently and securely is a growing challenge due to the increasing volume of video data being generated. In this paper, a video archive management method is presented to solve security and sensibility of the video files on blockchain. By encrypting video files and generating cryptographic hashes, the presented method can protect the integrity of the data and prevent unauthorized modifications. The method based on blockchain technology offers a compelling solution to the challenges faced in managing video archives. By combining the inherent security and transparency of blockchain with innovative data management algorithms, organizations can ensure the integrity, security, and accessibility of their video archives. © 2024 SPIE.

关键词： Data integrity

来源：评论

学校读者我要写书评

暂无评论

Efficient Homomorphic Approximation of Max Pooling for Privacy-Preserving Deep Learning 6th

Efficient Homomorphic Approximation of Max Pooling for Pri...

引用

6th International Conference on Machine Learning for Cyber Security, ML4CS 2024

作者： Zhang, Peng Qiu, Dongyan Duan, Ao Liu, Hongwei The Guangdong Key Laboratory of Intelligent Information Processing College of Electronics and Information Engineering Shenzhen University Guangdong Shenzhen518060 China College of Big Data and Internet Shenzhen Technology University Guangdong Shenzhen518118 China

ISBN: (纸本)9789819645657

Privacy-Preserving Deep Learning (PPDL) using Fully Homomorphic Encryption (FHE) addresses potential data privacy exposure risks associated with deploying deep learning models in untrusted cloud environments. FHE-based PPDL enables users to encrypt their data locally, allowing cloud service providers to perform computations directly on the encrypted data without ever accessing it. However, FHE faces challenges in efficiently handling nonlinear computations, which are essential to deep learning frameworks, particularly in implementing activation functions and max pooling layers. To address this challenge, we improve the homomorphic max pooling approximation scheme HMaxPool, and propose an efficient homomorphic max pooling algorithm, denoted as HMaxPool+. Specifically, we first redefine the approximation of the maximum function, and then design sub-algorithm to address the potential output expansion issue. The experimental results validate that both the homomorphic max pooling algorithm itself and the corresponding homomorphic SqueezeNet neural network demonstrate enhancements in computational efficiency and accuracy. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Differential privacy

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：