检索结果-内蒙古大学图书馆

48th IEEE International Conference on Acoustics, Speech and Signal processing, ICASSP 2023

作者： Xu, Mingbin Song, Congzheng Tian, Ye Agrawal, Neha Granqvist, Filip Van Dalen, Rogier Zhang, Xiao Argueta, Arturo Han, Shiyi Deng, Yaqiao Liu, Leo Walia, Anmol Jin, Alex Apple

ISBN: (纸本)9781728163277

Federated Learning (FL) is a technique to train models on distributed edge devices with local data samples. Differential Privacy (DP) can be applied with FL to provide a formal privacy guarantee for sensitive data on device. Our goal is to train a large neural network language model (NNLM) on compute-constrained devices while preserving privacy using FL and DP. However, the noise required to guarantee differential privacy increases as the model size grows, which often prevents convergence. We propose Partial Embedding Updates (PEU), a novel technique to reduce the impact of DP-noise by decreasing payload size. Furthermore, we adopt Low Rank Adaptation (LoRA) and Noise Contrastive Estimation (NCE) to reduce the memory demands of large models on compute-constrained devices. We demonstrate in simulation and with real devices that this combination of techniques makes it possible to train large-vocabulary language models while preserving accuracy and privacy. © 2023 IEEE.

关键词： Sensitive data

来源：评论

学校读者我要写书评

暂无评论

GIM (Ghost In the Machine): A Coarse -Grained Reconfigurable Compute -In -Memory Platform for Exploring Machine -Learning Architectures

GIM (Ghost In the Machine): A Coarse -Grained Reconfigurable...

引用

1st International Conference on Smart Energy Systems and Artificial Intelligence (SESAI)

作者： Borowicz, Maya Ding, James Fan, Winnie Gao, Zhongqi Jackson, Davis Lu, Ares Rohlfsen, Sophia Simar, Ray Rice Univ Houston TX 77005 USA

ISBN: (纸本)9798350364613;9798350364606

Machine-learning (ML) algorithms are finding wide adoption across a rich spectrum of application domains with diverse requirements in terms of performance, power, and cost. These diverse requirements are making it necessary to explore a large space of ML architectures and reexamine fundamental computational structures, a process of exploration that is very expensive. To get around the costly computations associated with large data sets and long training times, there have been increasing investments in specialized fixed-function hardware. However, this specialized hardware is expensive and hard to generalize to address the spectrum of applications. For our experiments, we focus on a novel, highly parallel, superset ML architecture, and use it to test the capabilities of new coarse-grained FPGAs containing hundreds and thousands of DSP slices with dedicated local storage. These new coarse-grained architectures allow us to achieve ASIC-like clock rate and reductions in power while exploring novel and common ML architectures.

关键词： AMD architecture back propagation coarse grained compute in memory DSP FPGA machine learning neural network parallel processing Rice University RISC V Simulink systolic UltraScale Zync

来源：评论

学校读者我要写书评

暂无评论

Cross-Attention Guided Group Aggregation network for Cropland Change Detection

引用

IEEE SENSORS JOURNAL 2023年第12期23卷 13680-13691页

作者： Xu, Chuan Ye, Zhaoyi Mei, Liye Shen, Sen Sun, Shaohua Wang, Ying Yang, Wei Hubei Univ Technol Sch Comp Sci Wuhan 430068 Peoples R China Naval Engn Univ Sch Weapon Engn Wuhan 430030 Peoples R China Wuchang Shouyi Univ Sch Informat Sci & Engn Wuhan 430032 Peoples R China

Cropland resources are essential for the provision of food production, which is one of the most fundamental needs of human life. Change detection (CD) technology enables the dynamic monitoring of high-resolution cropland resource images acquired through remote sensing satellite sensors. However, current CD methods are not capable of extracting meaningful change information from dense and continuously distributed cropland. In addition, the common feature fusion processing often results in information redundancy and the loss of key features. Therefore, we propose a cross-attention guided group aggregation network (CAGNet) to achieve effective cropland CD. Specifically, we adopt a cross-attention (CA) module to enhance the capability of extracting and characterizing the features of the changed region, reducing the influence of noise and pseudo-change on CD. To alleviate the loss of key information during the multiscale feature fusion process and thus improve the CD performance, we design a group aggregation (GA) module that gradually groups and aggregates the bitemporal features from coarse to fine. Finally, we use a fully convolutional network to obtain the detailed CD results. Furthermore, we demonstrate the effectiveness of knowledge transfer in the field of CD. It allows the models to obtain the underlying mechanisms and characterization capabilities of changed features on the building CD dataset in advance, which significantly improves the performance of various methods on the cropland CD dataset. The experimental results show that CAGNet's quantitative metrics results on the cropland dataset (CL-CD) outperform the other ten benchmarked methods, achieving an F1-score of 79.53%.

关键词： Feature extraction Sensors Remote sensing Data mining Knowledge transfer Production Convolutional neural networks Change detection (CD) cropland cross attention (CA) group aggregation (GA) remote sensing

来源：评论

学校读者我要写书评

暂无评论

A Hybrid Approach combining ANN-based and Conventional Demapping in Communication for Efficient FPGA-Implementation 36

A Hybrid Approach combining ANN-based and Conventional Demap...

引用

36th IEEE International Parallel and distributed processing Symposium (IEEE IPDPS)

作者： Ney, Jonas Hammoud, Bilal Wehn, Norbert TU Kaiserslautern Kaiserslautern Germany

ISBN: (纸本)9781665497473

In communication systems, Autoencoder (AE) refers to the concept of replacing parts of the transmitter and receiver by artificial neural networks (ANNs) to train the system end-to-end over a channel model. This approach aims to improve communication performance, especially for varying channel conditions, with the cost of high computational complexity for training and inference. Field-programmable gate arrays (FPGAs) have been shown to be a suitable platform for energy-efficient ANN implementation. However, the high number of operations and the large model size of ANNs limit the performance on the resource-constrained devices, which is critical for low-latency and high-throughput communication systems. To tackle his challenge, we propose a novel approach for efficient ANN-based demapping on FPGAs, which combines the adaptability of the AE with the efficiency of conventional demapping algorithms. After adaption to channel conditions, the channel characteristics, implicitly learned by the ANN, are extracted to enable the use of optimized conventional demapping algorithms for inference. We validate the hardware efficiency of our approach by providing FPGA implementation results and by comparing the communication performance to that of conventional systems. Our work opens a door for the practical application of ANN-based communication algorithms on FPGAs.

关键词： Training Performance evaluation Runtime Transmitters Artificial neural networks Receivers Inference algorithms

来源：评论

学校读者我要写书评

暂无评论

Research on Performance of Public Welfare Crowdfunding Based on neural network Intelligent Model

引用

SCIENTIFIC PROGRAMMING 2022年第0期2022卷

作者： Xia, Zhenqin Iowa State Univ Business Sch Ames IA 50010 USA

From the appearance of entrepreneurial narratives, it communicates the movement of innovatory suggestion and entrepreneurial madness on revenue deed in technology crowdfunding, outsearch the immediate percussion of innovatory hint on financier act, and the moderating sign of entrepreneurial emotion on the relationship between novelty and resources exploit. Through the mode of text-book analysis, the genuine technology design in JD crowdfunding was interested in bearing empiric trial of the mold. neural networks have large-scale parallel, distributed storage and processing, self-organization, self-adaptation, and self-learning capabilities and are especially suitable for dealing with inaccurate and ambiguous information processing problems that need to consider many factors and conditions at the same time. The empiric arise shows that forward-looking neologism queue indubitably moves financier completion, and primitive neology suggestion negatively pretends resources accomplishment;pierce People's entrepreneurial emotion intensify the dogmatic outcome of incremental neology humor on financier work, but has no momentous regulatory performance on the relationship between fundamental novation cue up and revenue exploit. On this base, a nerve plexure pattern is further fabricated for equal analysis. The characteristic efficiency of retrogradation example and nerval meshwork fashion is amended. In addition, based on the sample data of the Dahuotou fair crowdfunding platform, the indicators of the extended resource evaluation are judged based on the video revenue duration of the design scheme, scheme dynamics, extended scoring, non-hidden technology, and employee inclusiveness. There is an interaction manifestation between video and operant allowance data;the penetration sign of the two logical recession molds is emended than that of the multilayer McCulloch-Pitts neuron mold and the divergent base duty design.

关键词： Molds

来源：评论

学校读者我要写书评

暂无评论

Raw Ultrasound-Based Phonetic Segments Classification Via Mask Modeling 48

Raw Ultrasound-Based Phonetic Segments Classification Via Ma...

引用

48th IEEE International Conference on Acoustics, Speech and Signal processing, ICASSP 2023

作者： You, Kang Liu, Bo Xu, Kele Xiong, Yunsheng Xu, Qisheng Feng, Ming Csapó, Tamás Gábor Zhu, Boqing Tongji University Shanghai China National University of Defense Technology Changsha China National Key Lab of Parallel and Distributed Processing Changsha China Budapest University of Technology and Economics Budapest Hungary

ISBN: (纸本)9781728163277

Ultrasound tongue imaging is widely used in clinical linguistics and phonetics. Recently, deep neural networks, especially convolutional neural networks, have been widely used in the interpretation and analysis of ultrasound tongue images (UTI). Despite achieving satisfactory performance, deep models rely on a large amount of manually labeled data, which is often difficult to obtain in practical settings. To address this issue, this paper focuses on how to utilize a large amount of unlabeled UTI data to improve the performance of UTI classification task. Specifically, we explore self-supervised learning with masking modeling strategy. By predicting the masked part, our pre-trained model enables the neural network to infer contextual information. Then, we fine-tune the pre-trained model with a small amount of labeled data. Compared with the previous competing algorithms, our method can improve the classification accuracy by an average of 13.33% in four different scenarios. © 2023 IEEE.

关键词： Mask modeling Masked auto-encoder Self-supervised learning Ultrasound tongue imaging.

来源：评论

学校读者我要写书评

暂无评论

Hierarchical Collaborative Resource Scheduling in Industrial Internet of Things based on Graph neural networks and Deep Reinforcement Learning 29

Hierarchical Collaborative Resource Scheduling in Industrial...

引用

29th IEEE International Conference on Parallel and distributed Systems, ICPADS 2023

作者： Meng, Qifeng Luo, Zihui Zheng, Xiaolong Liu, Liang Ma, Huadong Beijing University of Posts and Telecommunications Beijing100876 China

ISBN: (纸本)9798350330717

The hierarchical cooperative resource scheduling architecture provides a promising direction for efficient collaborative processing of edge computing under the dynamic and intricate landscape of the Industrial Internet of Things (IIoT). However, existing scheduling algorithms often struggle to effectively capture the intricate information features inherent in hierarchical and collaborative domains, leading to suboptimal solutions. To tackle this challenge, we introduce a novel hierarchical cooperative resource scheduling framework based on Graph neural networks (GNN) and Deep Reinforcement Learning (DRL). We first leverage hierarchical GNN to facilitate seamless information exchange among internal nodes and adjacent nodes between layers in the hierarchical structure and transform it into node embeddings. These meticulously designed embeddings are then input into the policy model of DRL for the iterative learning process to generate higher-quality solutions by leveraging global feature information. Experiment results unequivocally demonstrate the superiority of our approach over baselines in terms of scheduling performance. Furthermore, our model exhibits robust generalization capabilities across various scenarios. © 2023 IEEE.

关键词： Deep Reinforcement Learning Edge Computing Graph neural network Hierarchical Collaborative Industrial Internet of Things

来源：评论

学校读者我要写书评

暂无评论

Multiscale Residual Pyramid network for Seismic Background Noise Attenuation

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2022年 60卷 1页

作者： Zhong, Tie Li, Feng Zhang, Rongzhe Dong, Xintong Lu, Shaoping Minist Educ Key Lab Modern Power Syst Simulat & Control & Ren Jilin 132012 Jilin Peoples R China Northeast Elect Power Univ Dept Commun Engn Coll Elect Engn Jilin 132012 Jilin Peoples R China Jilin Univ Coll Geo Explorat Sci & Technol Changchun 130026 Jilin Peoples R China Jilin Univ Coll Instrumentat & Elect Engn Changchun 130026 Jilin Peoples R China Southern Marine Sci & Engn Guangdong Lab Zhanjian Zhanjiang 524000 Guangdong Peoples R China Sun Yat Sen Univ Sch Earth Sci & Engn Guangzhou 510275 Guangdong Peoples R China Southern Marine Sci & Engn Guangdong Lab Zhuhai Zhuhai 519000 Guangdong Peoples R China Sun Yat Sen Univ Guangdong Prov Key Lab Geodynam & Geohazards Guangzhou 510275 Guangdong Peoples R China

Seismic background noise affects the recognition of reflection signals, thereby impeding the subsequent seismic data processing, such as seismic imaging and inversion. In addition, seismic background noise has relatively complex properties, such as nonstationarity and spectral aliasing, which can be further hampered with the deterioration of the exploration environment. Deep-learning methods have been successfully applied to effectively attenuate complex seismic noise and have shown significant improvements over conventional denoising methods. However, most denoising networks only utilize single-scale features, resulting in poor performance when confronted with seismic data in a low signal-to-noise ratio (SNR). To further enhance the denoising capability, a novel multiscale residual pyramid network (MRP-Net) was proposed to separate the desired signals and complex seismic noise. Compared with single-scale networks, MRP-Net can take advantage of the multiscale features, thereby improving noise attenuation capability. In general, the pyramid-like framework in MRP-Net can extract the potential features at different scales through downsampling and upsampling operations, and skip connections were applied to fuse the global-coarse and local-fine features. On this basis, a double-path spatial attention (DSA) module was designed to enhance the desired features, further improving the processing performance of separating the desired signals from the intense seismic background noise. Comprehensive experiments on synthetic and field seismic data demonstrate that MRP-Net is effective for complex seismic noise attenuation, both for conventional geophone-acquired data and distributed acoustic sensing (DAS) records.

关键词： Noise reduction Feature extraction Attenuation Filtering Convolution Data processing Convolutional neural networks Convolutional neural network (CNN) multiscale denoising network seismic noise attenuation spatial attention weak signal recovery

来源：评论

学校读者我要写书评

暂无评论

Application of Data Encryption in Chinese Named Entity Recognition 1

引用

32nd International Conference on Artificial neural networks (ICANN)

作者： Dong, Jikun Long, Kaifang Yu, Hui Xu, Weizhi Shandong Normal Univ Sch Informat Sci & Engn Jinan Peoples R China Shandong Normal Univ Sch Business Jinan Peoples R China Shandong Prov Key Lab Novel Distributed Comp Soft Jinan 250358 Peoples R China

ISBN: (数字)9783031441981

ISBN: (纸本)9783031441974;9783031441981

Recently, with the continuous development of deep learning, there has been a significant improvement in the performance of named entity recognition tasks. However, privacy and confidentiality concerns in specific fields, such as biomedical and military, limit the availability of data for training deep neural networks. To address the issues of data leakage and the disclosure of sensitive data in these domains, we propose an encryption learning framework. For the first time, we employ multiple encryption algorithms to encrypt the training data in the named entity recognition task, training the deep neural network with the encrypted data. Our experiments, conducted on six Chinese datasets, including three self-constructed datasets, demonstrate that the encryption method achieves satisfactory results. In fact, the performance of some models trained with encrypted data even surpasses that of the unencrypted method, highlighting the effectiveness of the introduced encryption method and partially resolving the problem of data leakage.

关键词： Chinese named entity recognition Data encryption Privacy protection Natural language processing Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Memory Efficient Deep neural network Training 27th

Memory Efficient Deep Neural Network Training

引用

27th International European Conference on Parallel and distributed Computing (Euro-Par)

作者： Shilova, Alena Inria Bordeaux Sud Ouest Bordeaux France Univ Bordeaux Bordeaux France

ISBN: (纸本)9783031061561;9783031061554

Recently Artificial Intelligence (AI) has demonstrated a huge progress in solving complex problems such as image classification, text generation, translation... Its success is due to a development of hardware and algorithms making possible the emergence of Deep neural networks (DNNs). Such DNNs are composed of a number of operations from numerical linear algebra, whose order is defined with a Directed Acyclic Graph (DAG). These DNNs operate at the limit of the computational resources, and to go to deeper and more complex neural networks it is necessary either to design more powerful computational resources or optimize their usage. In our work, we address in particular the problem of lowering memory usage during the training of DNNs.

关键词： neural networks Rematerialization Offloading Model parallelism

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：