检索结果-内蒙古大学图书馆

39th Annual AAAI Conference on artificial Intelligence, AAAI 2025

作者： Li, Teng Ma, Xingjun Jiang, Yu-Gang Shanghai Key Lab of Intell. Info. Processing School of CS Fudan University China

ISBN: (纸本)157735897X

Transferable adversarial examples highlight the vulnerability of deep neural networks (DNNs) to imperceptible perturbations across various real-world applications. While there have been notable advancements in untargeted transferable attacks, targeted transferable attacks remain a significant challenge. In this work, we focus on generative approaches for targeted transferable attacks. Current generative attacks focus on reducing overfitting to surrogate models and the source data domain, but they often overlook the importance of enhancing transferability through additional semantics. To address this issue, we introduce a novel plug-and-play module into the general generator architecture to enhance adversarial transferability. Specifically, we propose a Semantic Injection Module (SIM) that utilizes the semantics contained in an additional guiding image to improve transferability. The guiding image provides a simple yet effective method to incorporate target semantics from the target class to create targeted and highly transferable attacks. Additionally, we propose new loss formulations that can integrate the semantic injection module more effectively for both targeted and untargeted attacks. We conduct comprehensive experiments under both targeted and untargeted attack settings to demonstrate the efficacy of our proposed approach. © 2025, Association for the Advancement of artificial Intelligence (***). All rights reserved.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Accelerating End-to-End Deep Learning Workflow With Codesign of Data Preprocessing and Scheduling

引用

IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS 2021年第7期32卷 1802-1814页

作者： Cheng, Yang Li, Dan Guo, Zhiyuan Jiang, Binyao Geng, Jinkun Bai, Wei Wu, Jianping Xiong, Yongqiang Tsinghua Univ Beijing 100084 Peoples R China Beihang Univ Beijing 100191 Peoples R China Microsoft Res Beijing 100080 Peoples R China Shanghai Jiao Tong Univ Shanghai 200240 Peoples R China

In this article, we investigate the performance bottleneck of existing deep learning (DL) systems and propose DLBooster to improve the running efficiency of deploying DL applications on GPU clusters. At its core, DLBooster leverages two-level optimizations to boost the end-to-end DL workflow. On the one hand, DLBooster selectively offloads some key decoding workloads to FPGAs to provide high-performance online data preprocessing services to the computing engine. On the other hand, DLBooster reorganizes the computational workloads of training neural networks with the backpropagation algorithm and schedules them according to their dependencies to improve the utilization of GPUs at runtime. Based on our experiments, we demonstrate that compared with baselines, DLBooster can improve the image processing throughput by 1.4x - 2.5x and reduce the processing latency by 1/3 in several real-world DL applications and datasets. Moreover, DLBooster consumes less than 1 CPU core to manage FPGA devices at runtime, which is at least 90 percent less than the baselines in some cases. DLBooster shows its potential to accelerate DL workflows in the cloud.

关键词： Data preprocessing Training Field programmable gate arrays Task analysis Graphics processing units artificial neural networks Hardware Deep learning data preprocessing workload offloading computation scheduling FPGAs

来源：评论

学校读者我要写书评

暂无评论

Investigating the efficiency and performance gains of FPGA-accelerated Convolutional neural networks

Investigating the efficiency and performance gains of FPGA-a...

引用

作者： Shouyi Li Viterbi School of Engineering University of Southern California

With the rise of big data and artificial intelligence(AI),Convolutional neural networks(CNNs) have become instrumental in numerous applications,from image and speech recognition to natural language ***,these networks' computational demands often exceed the capabilities of traditional processing units,leading to a search for more effective computing *** research aims to evaluate the potential of Field-Programmable Gate Array(FPGA) technology in accelerating CNN computations,considering FPGA's unique attributes such as reprogrammability,energy efficiency,and custom logic *** primary aim of this research is to compare the efficiency and performance of FPGA acceleration of CNNs with conventional processing units like CPUs and GPUs and to explore its potential for future AI *** research employs a mixed-methods approach,including an integrated literature review and comparative *** paper reviews state-of-the-art research on FPGA-accelerated CNNs,benchmark performance metrics of FPGA,CPU,and GPU platforms across various CNN models,and compare FPGA-based AI applications with other real-world AI *** findings suggest a significant potential for FPGA-accelerated CNNs,particularly in scenarios requiring real-time computation or power-limited ***,challenges persist in the areas of development complexity and limited on-chip *** work must focus on surmounting these barriers to unlock the full potential of FPGAaccelerated CNNs.

关键词： CNN FPGA CPU GPU AI

来源：评论

学校读者我要写书评

暂无评论

Extreme learning machine versus classical feedforward network Comparison from the usability perspective

引用

neural COMPUTING & applications 2021年第22期33卷 15121-15144页

作者： Markowska-Kaczmar, Urszula Kosturek, Michal Wroclaw Univ Sci & Technol Wybrzeie Wyspianskiego 27 PL-50370 Wroclaw Poland

Our research is devoted to answering whether randomisation-based learning can be fully competitive with the classical feedforward neural networks trained using backpropagation algorithm for classification and regression tasks. We chose extreme learning as an example of randomisation-based networks. The models were evaluated in reference to training time and achieved efficiency. We conducted an extensive comparison of these two methods for various tasks in two scenarios: center dot using comparable network capacity and center dot using network architectures tuned for each model. The comparison was conducted on multiple datasets from public repositories and some artificial datasets created for this research. Overall, the experiments covered more than 50 datasets. Suitable statistical tests supported the results. They confirm that for relatively small datasets, extreme learning machines (ELM) are better than networks trained by the backpropagation algorithm. But for demanding image datasets, like imageNet, ELM is not competitive to modern networks trained by backpropagation;therefore, in order to properly address current practical needs in pattern recognition entirely, ELM needs further development. Based on our experience, we postulate to develop smart algorithms for the inverse matrix calculation, so that determining weights for challenging datasets becomes feasible and memory efficient. There is a need to create specific mechanisms to avoid keeping the whole dataset in memory to compute weights. These are the most problematic elements in ELM processing, establishing the main obstacle in the widespread ELM application.

关键词： Randomisation-based learning Extreme learning Backpropagation algorithm neural networks

来源：评论

学校读者我要写书评

暂无评论

Attack on Scene Flow Using Point Clouds

Attack on Scene Flow Using Point Clouds

引用

IEEE Workshop on Machine Learning for Signal processing

作者： Haniyeh Ehsani Oskouie Mohammad-Shahram Moin Shohreh Kasaei Sharif University of Technology ICT Research Institute

ISBN: (数字)9798350372250

ISBN: (纸本)9798350372267

Deep neural networks have made significant advancements in accurately estimating scene flow using point clouds, which is vital for many applications like video analysis, action recognition, and navigation. The robustness of these techniques, however, remains a concern, particularly in the face of adversarial attacks that have been proven to deceive state-of-the-art deep neural networks in many domains. Surprisingly, the robustness of scene flow networks against such attacks has not been thoroughly investigated. To address this problem, the proposed approach aims to bridge this gap by introducing adversarial white-box attacks specifically tailored for scene flow networks. Experimental results show that the generated adversarial examples obtain up to 33.7 relative degradation in average end-point error on the KITTI and FlyingThings3D datasets. The study also reveals the significant impact that attacks targeting point clouds in only one dimension or color channel have on average end-point error. Analyzing the success and failure of these attacks on the scene flow networks and their 2D optical flow network variants shows a higher vulnerability for the optical flow networks.

关键词： Point cloud compression image motion analysis Computer vision Navigation Estimation Authentication artificial neural networks Signal processing Robustness Optical flow

来源：评论

学校读者我要写书评

暂无评论

Atomically Thin Synaptic Devices for Optoelectronic Neuromorphic Vision

引用

ADVANCED MATERIALS TECHNOLOGIES 2023年第9期8卷

作者： Ahmed, Taimur Jannat, Azmira Krishnamurthi, Vaishnavi Aung, Thiha Mazumder, Aishani Zavabeti, Ali Syed, Nitu Daeneke, Torben Ou, Jian Zhen AI-Hourani, Akram Walia, Sumeet RMIT Univ Sch Engn Melbourne Vic 3001 Australia Pak Austria Fachhsch Inst Appl Sci & Technol Haripur 22620 Pakistan Australian Natl Univ Canberra Res Sch Phys Canberra ACT 2601 Australia Univ Melbourne Parkville Dept Chem Engn Melbourne Vic 3010 Australia

Imaging sensors with inbuilt processing capability are expected to form the backbone of low-latency and highly energy efficient artificial vision systems. A range of emerging atomically thin materials provide opportunities to exploit their electrical and optical properties for human vision and brain inspired functions. This work reports atomically thin nanosheets of beta-In2S3 which exhibit inherent persistent photoconductivity (PPC) under ultraviolet and visible wavelengths. This PPC effect enables beta-In2S3-based optoelectronic devices to optically mimic the dynamics of biological synapses. Based on the material characterizations, the PPC effect is attributed to the intrinsic defects in the synthesized beta-In2S3 nanosheet. Furthermore, the feasibility of adopting these atomically thin synaptic devices for optoelectronic neuromorphic hardware is demonstrated by implementing a convolutional neural network for image classification. As such, the demonstrated atomically thin nanosheets and optoelectronic synaptic devices provide a platform for scaling up complex vision-sensory neural networks, which can find many promising applications for multispectral imaging and neuromorphic computation.

关键词： artificial vision image sensor liquid metals neural networks optoelectronics

来源：评论

学校读者我要写书评

暂无评论

Enhancing neural Network Robustness Through Intelligent Data Transformation

Enhancing Neural Network Robustness Through Intelligent Data...

引用

IEEE Sensors applications Symposium, SAS

作者： Michael D. Ciocco Nidhal C. Bouaynaya Shreekanth Mandayam Electrical & Computer Engineering Rowan University Glassboro NJ USA Electrical Engineering Texas State University San Marcos TX USA

ISBN: (数字)9798350369250

ISBN: (纸本)9798350369267

This paper exploits the origins of artificial neural networks (ANN s) based on approximation theory, to refocus efforts on data transformation, and away from the exclusive attention on the development of more and more complex architectures. Machine learning (ML) networks efficiently accept raw input data and determine their own features for the training process. Features are chosen to optimize accuracy, but typically are not designed to be robust, leaving the network susceptible to noise. Thus, a case is made for the reintroduction of data transformation techniques, through an intelligent algorithm, to enhance the robustness of the ML networks. In a sense, the science and mathematics of AI algorithms must provide a path for the accurate, reliable, and rapid deployment of intelligent algorithms without an over-reliance of technological prowess such as inexpensive bandwidth and computing power. We developed an algorithm for image classification based on intelligent geometric transformations, which forms the input data for any neural network architecture of choice. Our results show that adopting intelligent data transformations for pre-processing network input leads, in comparison to a conventional raw image input convolutional neural network, to a more robust model to both random and adversarial noise, offering network designers enhanced control over the trade-off between accuracy and robustness.

关键词： Training Accuracy Noise Machine learning Computer architecture Robustness Vectors

来源：评论

学校读者我要写书评

暂无评论

Logic-Based neural Network for image Compression applications 3

Logic-Based Neural Network for Image Compression Application...

引用

3rd IEEE Middle East and North Africa Communications Conference (MENACOMM)

作者： Doughan, Ziad Kassem, Rola El-Hajj, Ahmad M. Haidar, Ali M. Beirut Arab Univ Dept Elect & Comp Engn Debbieh Lebanon

ISBN: (纸本)9781665434430

This paper presents a new lossy image compression technique using Logic-based Weightless neural networks, which underwrite two novel network architectures. The system endorses three processing phases, image optimization, inflation, and skimming. This research demonstrates an untraditional approach of auto-compression network guided by horizontal and vertical pixel intensity wavering trend. The performance of this new approach aligns with human's perception of singularities in a certain pattern. The potential of trend analysis in image compression incorporates with information storage techniques and knowledge accumulation. The weightless network models generate images underlying enough distinct features that preserve the originality of a particular pattern but give superior levels of compression.

关键词： Logic-Based Intelligence artificial neural networks Weightless neural networks image processing Compression

来源：评论

学校读者我要写书评

暂无评论

Chemometrics-based signal processing methods for biosensors in health and environment: A review

引用

ELECTROANALYSIS 2024年第7期36卷 e202300207-e202300207页

作者： Wu, Wanqing Yang, Jianlei Zhou, Yu Zheng, Qinggong Chen, Qing Bai, Zhaoao Niu, Jiaqi Dalian Maritime Univ Marine Engn Coll Dalian 116026 Peoples R China Engn Technol Ctr Ship Safety & Pollut Control Dalian 116026 Liaoning Peoples R China Liaoning Normal Univ Sch Life Sci Dalian 116026 Peoples R China Dalian Maritime Univ Marine Engn Coll Dalian 116026 Peoples R China

The increasing apprehension for health, safety and quality of life in modern society has resulted in the widespread use of biosensors. Biosensors are characterised by their high sensitivity, real-time monitoring, and easy integration, making them indispensable for environmental monitoring on-site, as well as invasive and non-invasive health monitoring. Signal processing and analysis are crucial to biosensor applications, with an important role being played by chemometrics in this regard. This review presents a review of recent research findings in the fields of environmental and health monitoring. In addition, it investigates the role that chemometrics plays in the processing and analysis of biosensor data. The research comprises conventional statistical techniques, including principal component analysis and wavelet transform, as well as modern techniques of artificial intelligence, such as machine learning with neural networks. Through the examination of various algorithm strengths and weaknesses, significant recommendations are offered for biosensor applications. Furthermore, the assessment delivers focused proposals for surmounting signal processing difficulties in biosensors. Additionally, the review contains a concise analysis and reflection on the issue of multiple detection and analysis. The review intends to give essential guidance to future researchers in selecting efficient and sensible methods of data processing for their studies. image

关键词： artificial intelligence algorithm biosensors chemometrics signal processing statistical algorithm

来源：评论

学校读者我要写书评

暂无评论

16th International Conference on Multi-disciplinary Trends in artificial Intelligence, MIWAI 2023

16th International Conference on Multi-disciplinary Trends i...

引用

16th International Conference on Multi-disciplinary Trends in artificial Intelligence, MIWAI 2023

ISBN: (纸本)9783031364013

The proceedings contain 72 papers. The special focus in this conference is on Multi-disciplinary Trends in artificial Intelligence. The topics include: Hybrid Model Using Interacted-ARIMA and ANN Models for Efficient Forecasting;demand and Price Forecasting Using Deep Learning Algorithms;conversational artificial Intelligence in Digital Healthcare: A Bibliometric Analysis;Statistical Analysis of the Monthly Costs of OPEC Crude Oil Using Machine Learning Models;Redefining the World of Medical image processing with AI – Automatic Clinical Report Generation to Support Doctors;sign Language Interpretation Using Deep Learning;interpretable Chronic Kidney Disease Risk Prediction from Clinical Data Using Machine Learning;traffic Prediction in Indian Cities from Twitter Data Using Deep Learning and Word Embedding Models;iAOI: An Eye Movement Based Deep Learning Model to Identify Areas of Interest;an Ensemble Technique to Detect Stress in Young Professional;automatic Differentiation Using Dual Numbers - Use Case;a Hybrid Intelligent Cryptography Algorithm for Distributed Big Data Storage in Cloud Computing Security;Low Light image Illumination Adjustment Using Fusion of MIRNet and Deep Illumination Curves;Multi-dimensional STAQR Indexing Algorithm for Drone applications;eduKrishnaa: A Career Guidance Web Application Based on Multi-intelligence Using Multiclass Classification Algorithm;iSTIMULI: Prescriptive Stimulus Design for Eye Movement Analysis of Patients with Parkinson’s Disease;AI Based Employee Attrition Prediction Tool;rescheduling Exams Within the Announced Tenure Using Reinforcement Learning;a Yolo-Based Deep Learning Approach for Vehicle Class Classification;pixel Value Prediction Task: Performance Comparison of Multi-Layer Perceptron and Radial Basis Function neural Network;sentiment Analysis of Twitter Data on ‘The Agnipath Yojana’;how Good are Transformers in Reordering?;LPCD: Incremental Approach for Dynamic networks.

关键词：

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：