检索结果-内蒙古大学图书馆

International Conference on Tools for Artificial Intelligence (ICTAI)

作者： Changyong Sun Tan Yang Youxun Lei State Key Laboratory of Networking and Switching Technology School of Computer Science (National Pilot Software Engineering School) Beijing University of Posts and Telecommunications Beijing China

Task scheduling and load balancing in heterogeneous computing environments has been a challenge for long, especially when dealing with multiple types of task input batches. In this scenario, existing methods cannot take into account both the high efficiency of task processing and the full utilization of cluster resources. However, the rise of artificial intelligence methods provides a new way to solve this problem. In this paper, we design a type-aware task scheduling method based on deep reinforcement learning to tackle multiple types of tasks in heterogeneous computing environment. First, we adopt prioritized dueling double deep q-learning network to make action decisions for each batch of input tasks. Then we build a task type prediction neural network to predict the task type of the input task, and then use the Monte Carlo algorithm based on reward value to realize the load balancing of the scheduled cluster. To verify the effectiveness of our proposed method, we use a widely used dataset Alibaba cluster trace dataset for our experiments. Experimental results show that our proposed algorithm can significantly shorten the average makespan of task batches and achieve better load balancing effect compared with other existing solutions.

关键词： Deep learning Schedules Q-learning Scheduling algorithms Clustering algorithms Load management Scheduling

来源：评论

学校读者我要写书评

暂无评论

Improving RGB-D Point Cloud Registration by Learning Multi-scale Local Linear Transformation

arXiv

引用

arXiv 2022年

作者： Wang, Ziming Huo, Xiaoliang Chen, Zhenghao Zhang, Jing Sheng, Lu Xu, Dong School of Software Beihang University China School of Electrical and Information Engineering The University of Sydney Australia Department of Computer Science The University of Hong Kong Hong Kong

Point cloud registration aims at estimating the geometric transformation between two point cloud scans, in which point-wise correspondence estimation is the key to its success. In addition to previous methods that seek correspondences by hand-crafted or learnt geometric features, recent point cloud registration methods have tried to apply RGB-D data to achieve more accurate correspondence. However, it is not trivial to effectively fuse the geometric and visual information from these two distinctive modalities, especially for the registration problem. In this work, we propose a new Geometry-Aware Visual Feature Extractor (GAVE) that employs multi-scale local linear transformation to progressively fuse these two modalities, where the geometric features from the depth data act as the geometry-dependent convolution kernels to transform the visual features from the RGB data. The resultant visual-geometric features are in canonical feature spaces with alleviated visual dissimilarity caused by geometric changes, by which more reliable correspondence can be achieved. The proposed GAVE module can be readily plugged into recent RGB-D point cloud registration framework. Extensive experiments on 3D Match and ScanNet demonstrate that our method outperforms the state-of-the-art point cloud registration methods even without correspondence or pose supervision. The code is available at https://***/514DNA/LLT. Copyright © 2022, The Authors. All rights reserved.

关键词： Linear transformations

来源：评论

学校读者我要写书评

暂无评论

BLOCK-TO-SCENE PRE-TRAINING FOR POINT CLOUD HYBRID-DOMAIN MASKED AUTOENCODERS

arXiv

引用

arXiv 2024年

作者： Zha, Yaohua Dai, Tai Wang, Yanzi Guo, Hang Zhang, Taolin Ouyang, Zhihao Fan, Chunlin Chen, Bin Chen, Ke Xia, Shu-Tao Tsinghua Shenzhen International Graduate School Tsinghua University China Institute of Visual Intelligence Pengcheng Laboratory China College of Computer Science and Software Engineering Shenzhen University China Harbin Institute of Technology Shenzhen China Aitist.ai

Point clouds, as a primary representation of 3D data, can be categorized into scene domain point clouds and object domain point clouds based on the modeled content. Masked autoencoders (MAE) have become the mainstream paradigm in point clouds self-supervised learning. However, existing MAE-based methods are domain-specific, limiting the model’s generalization. In this paper, we propose to pre-train a general Point cloud Hybrid-Domain Masked AutoEncoder (PointHDMAE) via a block-to-scene pre-training strategy. We first propose a hybrid-domain masked autoencoder consisting of an encoder and decoder belonging to the scene domain and object domain, respectively. The object domain encoder specializes in handling object point clouds and multiple shared object encoders assist the scene domain encoder in analyzing the scene point clouds. Furthermore, we propose a block-to-scene strategy to pre-train our hybrid-domain model. Specifically, we first randomly select point blocks within a scene and apply a set of transformations to convert each point block coordinates from the scene space to the object space. Then, we employ an object-level mask and reconstruction pipeline to recover the masked points of each block, enabling the object encoder to learn a universal object representation. Finally, we introduce a scene-level block position regression pipeline, which utilizes the blocks’ features in the object space to regress these blocks’ initial positions within the scene space, facilitating the learning of scene representations. Extensive experiments across different datasets and tasks demonstrate the generalization and superiority of our hybrid-domain model. The code will be released. Copyright © 2024, The Authors. All rights reserved.

关键词： Supervised learning

来源：评论

学校读者我要写书评

暂无评论

Beacon-based Indoor Fire Evacuation System using Augmented Reality and Machine Learning

Beacon-based Indoor Fire Evacuation System using Augmented R...

引用

IEEE International Conference on Robotic Computing (IRC)

作者： Hwawon Lee Dohyun Chung Seongmin Kim Jiwon Lim Yoonha Bahng Suhyun Park Anthony H. Smith School of Software Soongsil University Seoul Korea Industrial Security Chung-Ang University Seoul Korea Computer Science Kangwon National University Chuncheon Korea School of Software Kwangwoon University Seoul Korea School of Computer Science & Engineering Chung-Ang University Seoul Korea Computer Engineering Paichai University Seoul Korea Computer and Information Technology Purdue University West Lafayette United States

ISBN: (纸本)9781665472616

The inefficiency of fire evacuation has been the issue since the present evacuation method is unsuitable for complex buildings. In order to improve the evacuation system, this paper aims at three main components. First, Kalman Filter and deep learning models were utilized to estimate the user’s location accurately. Second, Q-learning based evacuation algorithm was designed to deal with various fire situations. Lastly, AR and a 2D map offer effective navigation systems. The proposed system offers the safest path based on accurate location with a user-friendly visual supplement.

关键词： Deep learning Location awareness Visualization Q-learning Machine learning algorithms Navigation Real-time systems

来源：评论

学校读者我要写书评

暂无评论

Survey on Image Object Detection Algorithms Based on Deep Learning 7th

Survey on Image Object Detection Algorithms Based on Deep Le...

引用

7th International Conference on Artificial Intelligence and Security, ICAIS 2021

作者： Fang, Wei Shen, Liang Chen, Yupeng School of Computer and Software Engineering Research Center of Digital Forensics Ministry of Education Nanjing University of Information Science and Technology Nanjing China Provincial Key Laboratory for Computer Information Processing Technology Soochow University Suzhou China

ISBN: (纸本)9783030786083

With the development of image processing technology, computer vision is becoming more and more popular. In recent years, deep learning has flourished, significant progress has been made in object detection. Especially after the R-CNN framework was proposed, the object detection framework based on deep learning has gradually become the mainstream, which can be divided into two categories: region-based and regression-based. Taking these two types of frameworks as the main body, this paper summarizes the research background and then discusses the object detection algorithms based on candidate regions represented by Faster R-CNN and the algorithms based on regression represented by the YOLO series. According to the development history, this paper summarizes the framework proposed in recent years, compares and analyzes the performance of object detection algorithms on public datasets, and introduces the application scenarios of those algorithms. Finally, this paper discusses the current difficulties and challenges in object detection and looks forward to the future development direction. © 2021, Springer Nature Switzerland AG.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

UT-GAN: A Novel Unpaired Textual-Attention Generative Adversarial Network for Low-Light Text Image Enhancement

UT-GAN: A Novel Unpaired Textual-Attention Generative Advers...

引用

IEEE International Conference on Image Processing

作者： Minglong Xue Zhengyang He Yanyi He Peiqi Xie Xin Feng School of Computer Science and Engineering Chongqing University of Technology Chongqing P.R. China State Key Laboratory for Novel Software Technology Nanjing University Nanjing P.R. China

How to balance lighting and texture details to achieve the desired visual effect remains the bottleneck of existing low-light image enhancement methods. In this paper, we propose a novel Unpaired Textual-attention Generative Adversarial N network (UT-GAN) for low-light text image enhancement task. UT-GAN first uses the Zero-DCE net for initial illumination recovery and our TAM module is proposed to translate text information into a textual attention mechanism for the overall network, emphasizing attention to the details of text regions. Moreover, the method constructs an AGM-Net module to mitigate noise effects and fine-tune the illumination. Experiments show that UT-GAN outperforms existing methods in qualitative and quantitative evaluation on the widely used the low-light datasets LOL and SID.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Privacy-Preserving Secure Shared Nearest Neighbor Clustering Scheme in Internet of Things

SSRN

引用

SSRN 2023年

作者： Hu, Qiqi Zhang, Hanlin Sun, Jiakang Lin, Jie Xu, Hansong Kong, Fanyu College of Computer Science and Technology Qingdao University Qingdao266071 China College of Computer Science and Technology University of Chinese Academy of Sciences 101408 China School of Electronic and Information Engineering Xi’an Jiaotong University Xian710049 China School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University Shanghai200240 China School of Software Shandong University Jinan250100 China

Shared Nearest Neighbor (SNN) is a density-based clustering approach extensively employed in industrial Internet of Things applications, including pattern recognition, image analysis, and data mining. The efficiency of SNN clustering poses a significant concern for IoT devices with constrained computational abilities, particularly when dealing with large-scale data. An effective solution is to delegate the computationally complex aspects of clustering to a remote cloud server. However, some security challenge may appear since the cloud environment is often untrustworthy. In this article, we proposed a privacy-preserving cloud-aided scheme for resource-limited IoT devices to execute SNN clustering securely and efficiently. To protect the privacy, we introduced a novel matrix transformation technique for encoding original data before transmitting it to the cloud. The cloud server processes the encoded data and returns results to the devices. The results will be decoded by device side to obtain the actual results. Besides, we developed a robust verification method to detect the accuracy of cloud’s result with optimal probability 1. Furthermore, we conducted comprehensive theoretical analysis and experimental evaluation, which confirmed the correctness, efficiency, and security for our scheme. In the final post-outsourcing tests, SNN's most complex step's time cost dropped from 10.859s to 1.183s in client side. © 2023, The Authors. All rights reserved.

关键词： Cluster analysis

来源：评论

学校读者我要写书评

暂无评论

Diving into Continual Ultra-fine-grained Visual Categorization

Diving into Continual Ultra-fine-grained Visual Categorizati...

引用

Proceedings of the Digital Image Computing: Technqiues and Applications (DICTA)

作者： Pengcheng Zhang Xiaohan Yu Xiao Bai Jin Zheng Xiaoyu Wu Yongsheng Gao School of Computer Science and Engineering State Key Laboratory of Software Development Environment Jiangxi Research Institute Beihang University Beijing China Institute for Integrated and Intelligent Systems Griffith University Nathan Australia

Recent advance in ultra-fine-grained visual categorization (ultra-FGVC) has significantly boosted the capability of deep neural networks for ultra-FGVC tasks. However, building models for continually learning to recognize increasing ultra-fine-grained categories is still under-explored. This limits the application of ultra-FGVC techniques in real-world production. To this, we take the first attempt for continual ultra-FGVC. By evaluating existing continual learning methods on the constructed continual ultra-FGVC benchmark, we observe that the main bottleneck lies in the limited model plasticity for incrementally adapting to new tasks. This can be caused by excessive anti-forgetting constraints as the difficult ultra-FGVC task requires substantial update of parameters, and over-fitting on early tasks given that the ultra-fine-grained categories are with very few training samples. To tackle these problems, we propose a joint self-supervised learning and prompting model. The prompt-based continual learning framework offers proper anti-forgetting operation by fixed pretrained vision transformer and adaptive prompt selection. By jointly optimizing the learnable prompts with an adversarial self-supervised loss, the over-fitting on each continual learning task is mitigated. Extensive experiments demonstrate that the proposed method outperforms existing continual learning methods on the challenging continual ultra-FGVC problem.

关键词：

来源：评论

学校读者我要写书评

暂无评论

SemLog: A Semantics-based Approach for Anomaly Detection in Big Data System Logs

SemLog: A Semantics-based Approach for Anomaly Detection in ...

引用

International Conference on Parallel and Distributed Systems (ICPADS)

作者： Xiaoman Tan Ningning Han Siyang Lu Wu Chen Dongdong Wang College of Computer and Information Science College of Software Southwest University Chongqing China School of Computer and Information Technology Beijing Jiaotong University Beijing China Department of Civil Environmental and Construction Engineering University of Central Florida Orlando FL US

Syslog-based anomaly detection is crucial for protecting the systems from malicious attacks or malfunctions. System logs are semi-structured text messages printed by logging statements to record the system’s run-time status, involving rich semantic information. However, the existing BERT-based log anomaly detection method is based on the log key sequence, does not consider the semantics of the log data, and discards the variable part, resulting in a high rate of missed detection. In this paper, we propose SemLog, a self-supervised framework for log anomaly detection based on BERT. By incorporating log semantics and variables and employing multi-feature fusion, we mitigate the independent assumption issue in the Masked Language Modeling model. The experimental results on three benchmarks show that SemLog achieves high performance compared with the state-of-the-art approaches for anomaly detection.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Node Representation Learning in Graph via Node-to-Neighbourhood Mutual Information Maximization

arXiv

引用

arXiv 2022年

作者： Dong, Wei Wu, Junsheng Luo, Yi Ge, Zongyuan Wang, Peng School of Computer Science and Engineering Northwestern Polytechnical University China School of Software Northwestern Polytechnical University China Monash University Australia University of Wollongong Australia

The key towards learning informative node representations in graphs lies in how to gain contextual information from the neighbourhood. In this work, we present a simple-yet-effective self-supervised node representation learning strategy via directly maximizing the mutual information between the hidden representations of nodes and their neighbourhood, which can be theoretically justified by its link to graph smoothing. Following InfoNCE, our framework is optimized via a surrogate contrastive loss, where the positive selection underpins the quality and efficiency of representation learning. To this end, we propose a topology-aware positive sampling strategy, which samples positives from the neighbourhood by considering the structural dependencies between nodes and thus enables positive selection upfront. In the extreme case when only one positive is sampled, we fully avoid expensive neighbourhood aggregation. Our methods achieve promising performance on various node classification datasets. It is also worth mentioning by applying our loss function to MLP based node encoders, our methods can be orders of faster than existing solutions. Our codes and supplementary materials are available at https://***/dongwei156/n2n. Copyright © 2022, The Authors. All rights reserved.

关键词： Machine learning

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：