检索结果-内蒙古大学图书馆

28th International Conference on database Systems for Advanced Applications, DASFAA 2023

作者： Hu, Richa Li, Dan Ng, See-Kiong Zheng, Zibin School of Software Engineering Sun Yat-Sen University Zhuhai528478 China Institute of Data Science National University of Singapore Singapore117602 Singapore School of Computing National University of Singapore Singapore117417 Singapore

ISBN: (纸本)9783031306778

In the era of big data, numerous data measurements collected from all walks of life are playing important roles in various data mining applications. Not all data owners (or keepers) could develop feasible learning models for knowledge discovery’s sake. Oftentimes, the original data need to be passed to or shared with researchers or data scientists for better mining insights, especially in the medical, financial, and industrial fields. However, concerns about sensitivity and privacy limit the availability and completeness of shared (or passed) data and the quality of mining results. In this paper, we propose a novel Convolutional Bidirectional Generative Adversarial Networks (CB-GAN) framework to generate sensitive synthetic data. The Convolutional Neural Networks are utilized to capture the feature correlations of the original data, and the Generative Adversarial Networks with Autoencoders are combined to synthesize realistically distributed data. To demonstrate the feasibility of the model, we evaluated it from three aspects: how similar are the distributions of the synthetic data to the original data, how well can the synthetic data accomplish future data mining tasks, and how much sensitive information has been hidden. Various experimental results showed the superiority of the proposed method compared with the state-of-the-art methods. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

Protein Generation via Genome-scale Language Models with Bio-physical Scoring

Protein Generation via Genome-scale Language Models with Bio...

引用

2023 International Conference on High Performance computing, Network, Storage, and Analysis, SC Workshops 2023

作者： Dharuman, Gautham Ward, Logan Ma, Heng Setty, Priyanka V. Gokdemir, Ozan Foreman, Sam Emani, Murali Hippe, Kyle Brace, Alexander Keipert, Kristopher Gibbs, Thomas Foster, Ian Anandkumar, Anima Vishwanath, Venkatram Ramanathan, Arvind Data Science and Learning Division United States Argonne Leadership Computing Facility Argonne National Laboratory United States Computer Science Department University of Chicago United States Nvidia Inc. United States California Institute of Technology United States

ISBN: (纸本)9798400707858

Large language models (LLMs) trained on vast biological datasets can learn biological motifs and correlations across the evolutionary landscape of natural proteins. LLMs can then be used for de novo design of novel proteins with specific structures, functions, and physicochemical properties. We employ a pre-trained genome-scale language model that uses codons as tokens and integrate it into a workflow for targeted generation of sequences. Our framework suggests new gene sequences that are ranked for downstream evaluation by metrics that collectively capture extensive sequence-specific, biophysical, and biochemical properties. We demonstrate our integrated workflow to design novel variants of the enzyme, malate dehydrogenase (MDH), that exhibit more favorable activation energies than their natural counterparts (reduction of 4.01 kJ/mol) with sustained sequence generation rates of ~104/hr and simulation rates of ~102/hr on 64 nodes of Polaris with about 99.7 system utilization during the run. © 2023 ACM.

关键词： Physicochemical properties

来源：评论

学校读者我要写书评

暂无评论

Monochromatic Image Dehazing Using Enhanced Feature Extraction Techniques in Deep Learning

Monochromatic Image Dehazing Using Enhanced Feature Extracti...

引用

Artificial Intelligence and Knowledge Discovery in Concurrent Engineering (ICECONF), International Conference on

作者： Nisarg Doshi Sagar Bhavsar D Rajeswari R Srinivasan Department of Data Science and Business Systems School of Computing College of Engineering and Technology SRM Institute of Science and Technology Kattankulathur Department of Computing Technologies School of Computing College of Engineering and Technology SRM Institute of Science and Technology Kattankulathur

Images photographed in foggy weather usually have poor visibility. To mitigate this problem researchers have come up with various image dehazing techniques. Now, more than ever, high-quality images that can be used to glean maximum information from autonomous systems are in high demand. This research work uses different Deep Learning (DL) architectures to draw out essential details from the picture and localize the information recovered to reduce the haze from the picture. The paper investigates to remove the hazes from the dehazed images using DL techniques. The first task of this proposed work attempts three pre-processing techniques namely, air light estimation, contextual regularization and boundary constraint. The second task of this work is to identify the suitable DL model to extract clear images from dehazed images. Evaluation metrics are PSNR value and SSIM value are used to estimate the values of dehazed images compared with clear images. Experimental results proves that AOD-Net outperforms good result with respect to PSNR value.

关键词： Deep learning Measurement Autonomous systems Atmospheric modeling Estimation Feature extraction Knowledge discovery

来源：评论

学校读者我要写书评

暂无评论

Joint Compression and Deadline Optimization for Communication-Efficient Federated Edge Learning

Joint Compression and Deadline Optimization for Communicatio...

引用

2023 IEEE Globecom Workshops, GC Wkshps 2023

作者： Zhang, Maojun Cai, Zhijie Liu, Dongzhu Jin, Richeng Zhu, Guangxu Zhong, Caijun Shenzhen Research Institute of Big Data Shenzhen China School of Computing Science University of Glasgow United Kingdom College of Information Science and Electronic Engineering Zhejiang University Hangzhou China Peng Cheng Laboratory Guangdong Shenzhen China Guangdong Guangzhou China

ISBN: (纸本)9798350370218

The federated edge learning (FEEL) framework is a popular approach for privacy-preserving collaborative model training, where edge devices and the server exchange learning updates frequently. In this vein, FEEL can cause a heavy burden on the communication system due to the limited bandwidth and severe interference between devices. Moreover, under the common FEEL framework, the server needs to wait for the update from the slowest device before starting the aggregation process, leading to the straggler issue that causes prolonged communication time. In this paper, we propose to accelerate FEEL from two aspects: i.e., 1) performing data compression on the edge devices and 2) setting a deadline on the edge server to exclude the straggler devices. However, undesired gradient compression errors and transmission outage are introduced by the aforementioned operations respectively, affecting the convergence of FEEL as well. To address these practical issues, we formulate a training time minimization problem that optimizes the compression ratio and deadline. To this end, the impact of compression error and transmission outage on the overall training time are quantified through convergence analysis. Then, the formulated problem is simplified and solved in an alternating manner, based on which, the novel joint compression and deadline optimization (JCDO) algorithm is derived. Numerical experiments show that the proposed method can greatly accelerate the vanilla FedAVG algorithm, and outperforms the state-of-the-art schemes. © 2023 IEEE.

关键词：

来源：评论

学校读者我要写书评

暂无评论

CoKV: Optimizing KV Cache Allocation via Cooperative Game

arXiv

引用

arXiv 2025年

作者： Sun, Qiheng Zhang, Hongwei Xia, Haocheng Zhang, Jiayao Liu, Jinfei Ren, Kui The State Key Laboratory of Blockchain and Data Security Zhejiang University China Institute of Blockchain and Data China Security China Siebel School of Computing and Data Science University of Illinois Urbana-Champaign United States

Large language models (LLMs) have achieved remarkable success on various aspects of human life. However, one of the major challenges in deploying these models is the substantial memory consumption required to store key-value pairs (KV), which imposes significant resource demands. Recent research has focused on KV cache budget allocation, with several approaches proposing head-level budget distribution by evaluating the importance of individual attention heads. These methods, however, assess the importance of heads independently, overlooking their cooperative contributions within the model, which may result in a deviation from their true impact on model performance. In light of this limitation, we propose CoKV, a novel method that models the cooperation between heads in model inference as a cooperative game. By evaluating the contribution of each head within the cooperative game, CoKV can allocate the cache budget more effectively. Extensive experiments show that CoKV achieves state-of-the-art performance on the LongBench benchmark using LLama-3-8B-Instruct and Mistral-7B models. Code is provided in https://***/nawei1010/CoKV. Copyright © 2025, The Authors. All rights reserved.

关键词： Budget control

来源：评论

学校读者我要写书评

暂无评论

Leveraging anatomical constraints with uncertainty for pneumothorax segmentation

Health Care Science

引用

Health Care science 2024年第6期3卷 456-474页

作者： Han Yuan Chuan Hong Nguyen Tuan Anh Tran Xinxing Xu Nan Liu Centre for Quantitative Medicine Duke-NUS Medical SchoolSingapore Department of Biostatistics and Bioinformatics Duke UniversityDurhamNorth CarolinaUSA Department of Diagnostic Radiology Singapore General HospitalSingapore Institute of High Performance Computing Agency for ScienceTechnology and ResearchSingapore Programme in Health Services and Systems Research Duke-NUS Medical SchoolSingapore Institute of Data Science National University of SingaporeSingapore

Background:Pneumothorax is a medical emergency caused by the abnormal accumulation of air in the pleural space—the potential space between the lungs and chest *** 2D chest radiographs,pneumothorax occurs within the thoracic cavity and outside of the mediastinum,and we refer to this area as“lung+space.”While deep learning(DL)has increasingly been utilized to segment pneumothorax lesions in chest radiographs,many existing DL models employ an end-to-end *** models directly map chest radiographs to clinician-annotated lesion areas,often neglecting the vital domain knowl-edge that pneumothorax is inherently ***:We propose a novel approach that incorporates the lung+space as a constraint during DL model training for pneumothorax segmentation on 2D chest *** circumvent the need for additional annotations and to prevent potential label leakage on the target task,our method utilizes external datasets and an auxiliary task of lung *** approach generates a specific constraint of lung+space for each chest ***,we have incorporated a discriminator to eliminate unreliable constraints caused by the domain shift between the auxiliary and target ***:Our results demonstrated considerable improvements,with average performance gains of 4.6%,3.6%,and 3.3%regarding intersection over union,dice similarity coefficient,and Hausdorff *** results were con-sistent across six baseline models built on three architectures(U-Net,LinkNet,or PSPNet)and two backbones(VGG-11 or MobileOne-S0).We further con-ducted an ablation study to evaluate the contribution of each component in the proposed method and undertook several robustness studies on hyper-parameter selection to validate the stability of our ***:The integration of domain knowledge in DL models for medical applications has often been *** research underscores the significance of incorporating medical domain kno

关键词： constrained optimization deep transfer learning diagnostic radiology pneumothorax detection semantic segmentation

来源：评论

学校读者我要写书评

暂无评论

SSRN

引用

SSRN 2023年

作者： Huang, Lin-Chieh Tsai, Hung-Hsu Institute of Data Science & Information Computing National Chung Hsing University Taichung402 Taiwan

This paper proposes a facial expression translation framework, called Similarity-aware Generative Adversarial Network (SimaGAN), which is applied in fine-grained image translation for facial expression images. It encodes an image with two features, styles and content features, which represent detailed class-related information and spatial structure, respectively. Moreover, a novel representation learning method is proposed in the design of SimaGAN, which is called similarity aggregation (SA). In the proposed model, SA is used in maximizing the similarity between a set of style features and its corresponding set of label embeddings, meanwhile, in minimizing the relative similarity between the false-negative style features to learn the disentangle representation. Here, the co-occurrence discriminator is also developed in the design of SimaGAN. The benefits of utilizing the co-occurrence discriminator contain to capture the texture of the source image and to preserve the detailed texture after the translation. Experimental results show that the proposed method outperforms others existing methods consideration here. © 2023, The Authors. All rights reserved.

关键词： Generative adversarial networks

来源：评论

学校读者我要写书评

暂无评论

An Adaptive Intrusion Detection System for IoT Networks Using Deep Learning 5

An Adaptive Intrusion Detection System for IoT Networks Usin...

引用

5th IEEE International Conference on Advances in Electronics, Computers and Communications, ICAECC 2023

作者： Harshavardhanan, Pon Muneeswaran, V. Saravanan, D. Vijayakarthik, P. Natchadalingam, R. Reddy, Pundru Chandra Shaker School of Computing Science and Engineering VIT Bhopal University Madhya Pradesh Kothrikalan India Computer Science and Engineering R.L Jalappa Institute of Technology Karnataka Doddaballapur India CSE-Data Science R.L. Jalappa Institute of Technology Karnataka Bengaluru India School of Computer Science and Engineering Lovely Professional University Punjab Phagwara India

ISBN: (纸本)9798350305449

The rapid expansion of the Internet-of-Things (IoT) has never before attracted the attention of cybercriminals. The increasing prevalence of cyberattacks against IoT devices and intermediary communication channels lends credence to this assertion. If an IoT attack goes unnoticed for too long, it might disrupt services severely, costing money. In addition, it poses the risk of having one's identity stolen. For IoT-enabled services to be dependable, secure and financially fruitful, real-time intrusion detection on IoT devices is a must. A new intrusion detection design for IoT devices based on Deep Learning (DL) is presented in this research. In order to identify potentially harmful traffic that could lead to assaults on Internet of Things (IoT) gadgets, this smart framework employs a 4layer deep Fully-Connected (FC) network architecture. To simplify deployment, the suggested system was designed to work with any available communication protocol. During experimental performance evaluation, the suggested system has proven to be effective against both simulated and real invasions. According to the data, the proposed model is more accurate than existing methods. This novel deep learning-based IDS has highest detection rate of 95.7 percent, making it suitable for bolstering the safety of IoT networks. © 2023 IEEE.

关键词： Internet of things

来源：评论

学校读者我要写书评

暂无评论

Improving Time-Series Classification Accuracy Based on Temporal Feature Representation Learning Using CRU-LSTM Autoencoder 5

Improving Time-Series Classification Accuracy Based on Tempo...

引用

5th IEEE International Conference on Cognitive Machine Intelligence, CogMI 2023

作者： Kim, Dohee Sim, Sunghyun Yoon, Bori Liu, Ling Bae, Hyerim Pusan National University Major in Industrial Data Science & Engineering Dept. Industrial Engineering Busan Korea Republic of Dong-Eui University Dept. Industrial Management & Big Data Engineering Busan Korea Republic of College of Computing Georgia Institute of Technology Georgia United States

ISBN: (纸本)9798350323832

Time-series data consists of a sequence of observations recorded in chronological order, where the data changes over time. This type of data exhibits various characteristics, such as temporal volatility, trends, and seasonality. Recently, a new layer structure called Correlation Recurrent Units (CRU) has been proposed to capture not only temporal variability but also trends and seasonality in time-series data. In this study, we propose an end-to-end model that utilizes a CRU autoencoder to learn temporal feature representations and address time-series classification problems simultaneously. To validate the performance of our proposed model, we conducted comparative experiments using 30 time-series classification datasets from six different types. The experimental results showed that the proposed model outperformed the baseline models in 20 out of the 30 time-series datasets. This indicates that the proposed approach effectively captures various temporal features in time-series data and improves the performance of time-series classification tasks. © 2023 IEEE.

关键词： Time series

来源：评论

学校读者我要写书评

暂无评论

Digital Forensics Analysis of iMazing Phone Explorer Tool to Aid Mobile Forensics Investigation

Digital Forensics Analysis of iMazing Phone Explorer Tool to...

引用

International Symposium on Digital Forensic and Security (ISDFS)

作者： Ashar Neyaz Avinash Kumar Bingyu Liu School of Computing & Data Science Wentworth Institute of Technology Boston MA USA Department of Computer Science Sam Houston State University Huntsville TX USA

ISBN: (数字)9798331509934

ISBN: (纸本)9798331509941

Apple devices continue to lead the smart device industry, with over 2.2 billion active devices worldwide. These devices, including iPhones and iPads, store extensive personally identifiable information (PII), making them vital in digital forensic investigations. This study specifically examines the retrieval and analysis of PII from an iPhone and an iPad using iMazing software as the sole forensic tool. data acquisition was conducted via a Lightning-to-VSB-A connection without jailbreaking, focusing on two distinct scenarios-one for each device-restricting access to user-space data. The analysis uncovered a range of forensic artifacts, including iCloud-synced SSID data, message logs, Wi-Fi security protocols, and other critical digital traces. Extracted data appeared in multiple formats, such as text files, CSV files, and plist files, providing forensic investigators with valuable insights. While this study is limited to iMazing, its findings highlight the potential and limitations of third-party tools in mobile forensics. By leveraging iMazing's capabilities, this research contributes to the evolving field of mobile device forensics, reinforcing the role of Apple devices in enhancing digital investigative methodologies.

关键词： data privacy Protocols Digital forensics Tablet computers Software Mobile handsets Security Smart devices Wireless fidelity Identification of persons

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：