检索结果-内蒙古大学图书馆

22nd Mexican International conference on artificial Intelligence (MICAI)

作者： Villegas-Cortez, Juan Roman-Alonso, Graciela Fernandez De Vega, Francisco Flores-Morales, Yafte Aaron Cordero-Sanchez, Salomon Univ Autonoma Metropolitana Dept Sistemas Azapotzalco Av Sn Pablo 180 Mexico City 02200 DF Mexico Univ Autonoma Metropolitana Dept Ingn Elect Unidad Iztapalapa Mexico City DF Mexico Univ Extremadura Dept Comp Sci C Santa Teresa Jornet38 Merida 06800 Spain Univ Autonoma Metropolitana Unidad Iztapalapa Dept Chem Mexico City DF Mexico

ISBN: (纸本)9783031477645;9783031477652

Pattern recognition has been evolving to include problems posed by new sceneries containing a high number of pattern components. processing this volume of information allows a more exact classification in wider types of applications;however, some of the difficulties of this scheme is the maintenance of numerical precision and mainly the reduction of the execution time. During the last 15 years, several Machine Learning solutions have been implemented to reduce the number of pattern components to be analyzed, such as artificial neural networks. Deep learning is an appropriate tool to accomplish this task. In this paper, a convolutional neural network is implemented for recognition and classification of human activity signals and digital images. It is achieved by automatically adjusting the parameters of the neural network through genetic algorithms using a multiprocessor and GPU platform. The results obtained show the reduction of computational costs and the possibility of better understanding of the solutions provided by Deep Learning.

关键词： Deep Learning Pattern recognition HAR image Recognition Parallel Genetic Algorithms

来源：评论

学校读者我要写书评

暂无评论

S-E Pipeline: A Vision Transformer (ViT) based Resilient Classification Pipeline for Medical Imaging Against Adversarial Attacks

S-E Pipeline: A Vision Transformer (ViT) based Resilient Cla...

引用

International Joint conference on neural networks (IJCNN)

作者： Neha, A. S. Chaturvedi, Vivek Shafique, Muhammad Indian Inst Technol Palakkad Dept CSE Kanjikode Kerala India New York Univ Div Engn Abu Dhabi U Arab Emirates

ISBN: (数字)9798350359312

ISBN: (纸本)9798350359329;9798350359312

Vision Transformer (ViT) is becoming widely popular in automating accurate disease diagnosis in medical imaging owing to its robust self-attention mechanism. However, ViTs remain vulnerable to adversarial attacks that may thwart the diagnosis process by leading it to intentional misclassification of critical disease. In this paper, we propose a novel image classification pipeline, namely, S-E Pipeline, that performs multiple pre-processing steps that allow ViT to be trained on critical features so as to reduce the impact of input perturbations by adversaries. Our method uses a combination of segmentation and image enhancement techniques such as Contrast Limited Adaptive Histogram Equalization (CLAHE), Unsharp Masking (UM), and High-Frequency Emphasis filtering (HFE) as preprocessing steps to identify critical features that remain intact even after adversarial perturbations. The experimental study demonstrates that our novel pipeline helps in reducing the effect of adversarial attacks by 72.22% for the ViT-b32 model and 86.58% for the ViT-l32 model. Furthermore, we have shown an end-to-end deployment of our proposed method on the NVIDIA Jetson Orin Nano board to demonstrate its practical use case in modern hand-held devices that are usually resource-constrained.

关键词： vision transformers medical imaging adversarial attacks defense mechanisms image enhancement segmentation

来源：评论

学校读者我要写书评

暂无评论

FPGA-Based Lightweight CNN Acceleration System for Real-Time Recognition 9

FPGA-Based Lightweight CNN Acceleration System for Real-Time...

引用

9th International conference on Integrated Circuits and Microsystems, ICICM 2024

作者： Ju, Junlei Liu, Cheng School of Microelectronics Shanghai University Shanghai China

ISBN: (纸本)9798331509453

In recent years, convolutional neural networks (CNNs) have become the core of many artificial intelligence applications, especially in fields such as image recognition and speech recognition. Deploying convolutional neural networks in hardware, as opposed to software, can increase speed and reduce power consumption. In this article, we propose an FPGA-based convolutional neural network acceleration system. This system optimizes LeNet-5 into a lightweight convolutional neural network model by replacing traditional convolution with depthwise separable convolutions and reducing the number of fully connected layers. After designing a parallel processing scheme for the computation process of the model, a CNN acceleration IP core is implemented using Verilog and applied to real-time handwritten digit recognition. This system can recognize one frame of image in 326.24μs, which is approximately 1s faster than CPU recognition. The total power consumption of the entire system is 1.947W, meeting the requirements of high real-time performance and low power consumption. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Lane Detection Method Based on Deformable Linear Convolution

Lane Detection Method Based on Deformable Linear Convolution

引用

International Joint conference on neural networks (IJCNN)

作者： Zhu Yuchang Xiao Nanfeng South China Univ Techol Sch Comp Sci & Eng Guangzhou 510006 Peoples R China

ISBN: (数字)9798350359312

ISBN: (纸本)9798350359329;9798350359312

Autonomous driving systems mainly rely on accurate detection of lane markings for navigation and safety. This paper explores an enhanced lane detection methodology employing deformable linear convolution, which dynamically adjusts to geometric variations of road markings. Our method aims to improve the detection fidelity under a range of challenging conditions such as variable illumination, road wear, and diverse weather scenarios, as evidenced by our experiments on the BDD100K dataset. The results demonstrate an improvement over those traditional lane detection techniques, suggesting that the deformable linear convolution offers a viable path forward for complex environmental adaptation in real-time image processing. Nevertheless, the computational demands of the proposed method in this paper highlight an area for further optimization. This study contributes to the field by providing an adaptable framework for the lane detection and sets the stage for future research focused on operational efficiency.

关键词： Lane Detection Semantic Segmentation Deformable Convolution Autonomous Driving neural networks

来源：评论

学校读者我要写书评

暂无评论

Efficient Stage Features for Edge Detection

Efficient Stage Features for Edge Detection

引用

9th International conference on Signal and image processing (ICSIP)

作者： Ji, Shucheng Yuan, Xiaochen Bao, Junqi Macao Polytech Univ Fac Appl Sci Macau Macao Peoples R China

ISBN: (纸本)9798350350920

Edge detection is a fundamental task in machine vision that facilitates feature extraction and representation across various visual domains, such as panoptic segmentation, autonomous driving, and image recognition. Despite the superior performance of current neural network-based edge detectors, the large parameter size renders edge detection models unsuitable for direct application in complex scenarios. Consequently, designing a compact edge detection network remains an imperative challenge. In this paper, we introduce the Efficient Stage Features Edge Detector (ESFED), a low-parameter, high-performance edge detector. ESFED is primarily composed of an efficient stage feature extractor, an upsampling network for edge features, and a feature fusion network for prediction, totaling only 51K parameters. It achieves 0.829 Optimal Dataset Scale (ODS) and 0.846 Optimal image Scale (OIS) on the Unified Dataset for Edge Detection (UDED) dataset, demonstrating notable performance in comparison to other state-of-the-art models.

关键词： Edge detection Deep neural networks Deep Learning

来源：评论

学校读者我要写书评

暂无评论

Chemical language models for molecular design

引用

MOLECULAR INFORMATICS 2024年第1期43卷 e202300288-e202300288页

作者： Bajorath, Juergen Rheinische Friedrich Wilhelms Univ Bonn Bonn Aachen Int Ctr Informat Technol Dept Life Sci Informat Bonn Germany Rheinische Friedrich Wilhelms Univ Bonn Lamarr Inst Machine Learning & Artificial Intellig Bonn Germany Rheinische Friedrich Wilhelms Univ Bonn Bonn Aachen Int Ctr Informat Technol Dept Life Sci Informat Friedrich Hirzebruch Allee 5-6 D-53115 Bonn Germany

In drug discovery, chemical language models (CLMs) originating from natural language processing offer new opportunities for molecular design. CLMs have been developed using recurrent neural network (RNN) or transformer architectures. For the predictive performance of RNN-based encoder-decoder frameworks and transformers, attention mechanisms play a central role. Among others, emerging application areas for CLMs include constrained generative modeling and the prediction of chemical reactions or drug-target interactions. Since CLMs are applicable to any compound or target data that can be presented in a sequential format and tokenized, mappings of different types of sequences can be learned. For example, active compounds can be predicted from protein sequence motifs. Novel off-the-beat-path applications can also be considered. For example, analogue series from medicinal chemistry can be perceived and represented as chemical sequences and extended with new compounds using CLMs. Herein, methodological features of CLMs and different applications are discussed. image

关键词： drug design language models recurrent neural networks encoder-decoder frameworks transformers attention mechanisms

来源：评论

学校读者我要写书评

暂无评论

Tensor-Based Chaotic Convolutional neural Network for Remote Sensing Data Classification 2

Tensor-Based Chaotic Convolutional Neural Network for Remote...

引用

2nd IEEE International conference on Signal, Information and Data processing, ICSIDP 2024

作者： Chen, Luobing Yin, Junjun Yang, Jian University of Science and Technology Beijing Beijing China Tsinghua University Beijing China

ISBN: (纸本)9798331515669

With the advancement of deep learning techniques, the classification of remote sensing data using artificial neural networks has emerged as a prominent research area. Despite this progress, the emulation of brain structures by traditional artificial neural networks remains at a relatively low level. Therefore, there is an urgent need to explore novel methods that can more effectively simulate human brain functions and enhance algorithmic performance. At the same time, artificial neural networks in image processing require the transformation of raw data into vector form. However, this conversion disrupts the inherent relationships between adjacent positional data in the raw data, resulting in the loss of crucial spatial structural information. To address these issues, we propose a tensor-based chaotic convolutional neural network model for the classification of remote sensing data. Firstly, we constructed a 3D discrete chaotic system, utilizing both Logistic mapping and Tent mapping, to achieve a more uniform iterative distribution and a broader full mapping range. This system was then integrated with convolutional neural networks to devise a novel chaotic convolutional neural network algorithm. By introducing chaotic mechanisms, this algorithm addresses the drawback of neural networks being prone to local minima. Secondly, we established a tensor model for remote sensing data. In contrast to existing methods that convert raw data into vector form, our approach represents raw data in tensor form, thereby preserving the spatial structural information among the three channels of the raw data. Finally, the effectiveness of the algorithm was validated using the NWPU-RESISC45 remote sensing dataset © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Multiscale Low-Frequency Memory Network for Improved Feature Extraction in Convolutional neural networks 38

Multiscale Low-Frequency Memory Network for Improved Feature...

引用

38th AAAI conference on artificial Intelligence (AAAI) / 36th conference on Innovative applications of artificial Intelligence / 14th Symposium on Educational Advances in artificial Intelligence

作者： Wu, Fuzhi Wu, Jiasong Kong, Youyong Yang, Chunfeng Yang, Guanyu Shu, Huazhong Carrault, Guy Senhadji, Lotfi Southeast Univ Key Lab New Generat Artificial Intelligence Techn Nanjing Jiangsu Peoples R China Univ Rennes Lab Traitement Signal & Image Rennes France Ctr Rech Informat Biomed Sinofrancais CRIBs Rennes France Southeast Univ Jiangsu Prov Joint Int Res Lab Med Informat Proc Nanjing Jiangsu Peoples R China

ISBN: (纸本)1577358872

Deep learning and Convolutional neural networks (CNNs) have driven major transformations in diverse research areas. However, their limitations in handling low-frequency information present obstacles in certain tasks like interpreting global structures or managing smooth transition images. Despite the promising performance of transformer structures in numerous tasks, their intricate optimization complexities highlight the persistent need for refined CNN enhancements using limited resources. Responding to these complexities, we introduce a novel framework, the Multiscale Low-Frequency Memory (MLFM) Network, with the goal to harness the full potential of CNNs while keeping their complexity unchanged. The MLFM efficiently preserves low-frequency information, enhancing performance in targeted computer vision tasks. Central to our MLFM is the Low-Frequency Memory Unit (LFMU), which stores various low-frequency data and forms a parallel channel to the core network. A key advantage of MLFM is its seamless compatibility with various prevalent networks, requiring no alterations to their original core structure. Testing on imageNet demonstrated substantial accuracy improvements in multiple 2D CNNs, including ResNet, MobileNet, EfficientNet, and ConvNeXt. Furthermore, we showcase MLFM's versatility beyond traditional image classification by successfully integrating it into image-to-image translation tasks, specifically in semantic segmentation networks like FCN and U-Net. In conclusion, our work signifies a pivotal stride in the journey of optimizing the efficacy and efficiency of CNNs with limited resources. This research builds upon the existing CNN foundations and paves the way for future advancements in computer vision. Our codes are available at https://***/AlphaWuSeu/MLFM.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

Quantitative comparison of the computational complexity of optical, digital and hybrid neural network architectures for image classification tasks

引用

OPTICS EXPRESS 2023年第26期31卷 44474-44485页

作者： Chen, Mengxiang Schoenhardt, Steffen Gu, Min Goi, Elena Univ Shanghai Sci & Technol Inst Photon Chips Shanghai 200093 Peoples R China Univ Shanghai Sci & Technol Ctr Artificial Intelligence Nanophoton Sch Opt Elect & Comp Engn Shanghai 200093 Peoples R China

By implementing neuromorphic paradigms in processing visual information, machine learning became crucial in an ever-increasing number of applications of our everyday lives, ever more performing but also computationally demanding. While a pre-processing of the information passively in the optical domain, before optical-electronic conversion, can reduce the computational requirements for a machine learning task, a comprehensive analysis of computational requirements for hybrid optical-digital neural networks is thus far missing. In this work we critically compare and analyze the performance of different optical, digital and hybrid neural network architectures with respect to their classification accuracy and computational requirements for analog classification tasks of different complexity. We show that certain hybrid architectures exhibit a reduction of computational requirements of a factor >10 while maintaining their performance. This may inspire a new generation of co-designed optical-digital neural network architectures, aimed for applications that require low power consumption like remote sensing devices.

关键词： Machine learning neural networks Optical neural systems Optical systems Point spread function Spatial light modulators

来源：评论

学校读者我要写书评

暂无评论

Towards Transferable Adversarial Attacks with Centralized Perturbation 38

Towards Transferable Adversarial Attacks with Centralized Pe...

引用

38th AAAI conference on artificial Intelligence (AAAI) / 36th conference on Innovative applications of artificial Intelligence / 14th Symposium on Educational Advances in artificial Intelligence

作者： Wu, Shangbo Tan, Yu-an Wang, Yajie Ma, Ruinan Ma, Wencong Li, Yuanzhang Beijing Inst Technol Sch Cyberspace Sci & Technol Beijing Peoples R China Beijing Inst Technol Sch Comp Sci & Technol Beijing Peoples R China

ISBN: (纸本)1577358872

Adversarial transferability enables black-box attacks on unknown victim deep neural networks (DNNs), rendering attacks viable in real-world scenarios. Current transferable attacks create adversarial perturbation over the entire image, resulting in excessive noise that overfit the source model. Concentrating perturbation to dominant image regions that are model-agnostic is crucial to improving adversarial efficacy. However, limiting perturbation to local regions in the spatial domain proves inadequate in augmenting transferability. To this end, we propose a transferable adversarial attack with fine-grained perturbation optimization in the frequency domain, creating centralized perturbation. We devise a systematic pipeline to dynamically constrain perturbation optimization to dominant frequency coefficients. The constraint is optimized in parallel at each iteration, ensuring the directional alignment of perturbation optimization with model prediction. Our approach allows us to centralize perturbation towards sample-specific important frequency features, which are shared by DNNs, effectively mitigating source model overfitting. Experiments demonstrate that by dynamically centralizing perturbation on dominating frequency coefficients, crafted adversarial examples exhibit stronger transferability, and allowing them to bypass various defenses.

关键词： Deep neural networks

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：