检索结果-内蒙古大学图书馆

Joint Source-Channel Coding for a Multivariate Gaussian Over a Gaussian MAC Using variational Domain Adaptation

IEEE TRANSACTIONS ON COGNITIVE COMMUNICATIONS AND NETWORKING 2023年第6期9卷 1424-1437页

作者： Li, Yishen Chen, Xuechen Deng, Xiaoheng Cent South Univ Sch Comp Sci & Engn Changsha 410083 Peoples R China

With the development of the distributed learning and edge computing, servers must often receive information from multiple terminal devices;thus, the importance of source-channel coding for distributed sources over multiple access channels (MACs) becomes more and more significant. This letter presents a deep joint source-channel coding (JSCC) design for a multivariate Gaussian source over a Gaussian MAC. The widely used autoencoder based deep-JSCC cannot perform stably under such conditions due to their easiness to fall into local optimum. Therefore we propose the variational domain adaptation (VDA)-JSCC scheme. Firstly, the loss function with an additional regularization term is introduced through variational analysis. The crucial prior distribution related to this item is obtained by domain adaptation, which is a transfer learning method. The proposed fine-tuning technique during the training process yields further performance improvement. Experiment results show that VDA-JSCC can always learn reasonable coding structures without artificial design and outperforms other state-of-the-art methods under different channel signal-to-noise ratios (CSNRs). We have also analyzed the reason why the performance of VDA-JSCC deteriorates in high CSNR range and then replace the encoder of VDA-JSCC with Mixture-of-Experts to improve its performance in high CSNR range. Finally, VDA-JSCC exhibits considerable robustness when the channel quality or correlation coefficient varies.

关键词： Encoding Symbols Feature extraction Decoding Standards Delays Adaptation models Deep joint source-channel coding domain adaptation variational autoencoder multivariate Gaussian multiple-access channel

来源：评论

学校读者我要写书评

暂无评论

Fast marching method assisted permeability upscaling using a hybrid deep learning method coupled with particle swarm optimization

GEOENERGY SCIENCE AND ENGINEERING

引用

GEOENERGY SCIENCE AND ENGINEERING 2023年 230卷

作者： Yousefzadeh, Reza Ahmadi, Mohammad Amirkabir Univ Technol Dept Petr Engn Tehran Iran

Geo-cellar models contain millions of gridblocks and are very time-consuming to simulate. This challenge necessitates upscaling geo-cellular models to obtain fit-for-purpose models for simulation. Simulation-based upscaling methods are computationally demanding, especially if there is a large number of geological realizations. On the other hand, static methods do not take into account the dynamics of the flow. In this paper, a novel hybrid deep learning-based upscaling method is proposed that can handle highly channelized and layered reservoir models. The proposed method combines the ConvLSTM network with the fast marching method (FMM), which is a computationally efficient method to get the dynamic response of the reservoir without the need for full-physics flow simulation. Since reservoir models have deposited during different geological periods, they have a temporal distribution along the third dimension. ConvLSTM layers can simultaneously extract spatiotemporal dependencies that other networks in the literature were not able to do so. The proposed method follows a twostep training process. In the first step, a variational autoencoder with ConvLSTM blocks (ConvLSTM-VAE) is trained and validated unsupervised using 10,000 permeability realizations. This network learns the datagenerating distribution of the realizations and improves the overall upscaling process. Then, the acquired weights are used to initialize the ConvLSTM-FMM network, which is fine-tuned by the particle swarm optimization to accomplish the upscaling task by minimizing the difference between the pressure drop of the fine-scale and upscaled realizations calculated by the FMM. Validity of the FMM to estimate the BHP drop of the wells was verified on the utilized reservoir model. The advantage of the proposed data-driven method is that it can be trained once and used to upscale new realizations, contrary to conventional simulation-based methods that require running the entire process whenever the fine

关键词： Upscaling Deep learning variational autoencoder Fast marching method Unsupervised

来源：评论

学校读者我要写书评

暂无评论

Joint Graph Attention and Asymmetric Convolutional Neural Network for Deep Image Compression

引用

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY 2023年第1期33卷 421-433页

作者： Tang, Zhisen Wang, Hanli Yi, Xiaokai Zhang, Yun Kwong, Sam Kuo, C. -C. Jay Tongji Univ Dept Comp Sci & Technol Shanghai 201804 Peoples R China Tongji Univ Serv Comp Minist Educ Shanghai 200092 Peoples R China Tongji Univ Key Lab Embedded Syst Shanghai 200092 Peoples R China Chinese Acad Sci Shenzhen Inst Adv Technol Shenzhen 518055 Peoples R China City Univ Hong Kong Dept Comp Sci Hong Kong Peoples R China Univ Southern Calif Los Angeles Signal & Image Proc Inst Ming Hsieh Dept Elect Engn Los Angeles CA 90089 USA

Recent deep image compression methods have achieved prominent progress by using nonlinear modeling and powerful representation capabilities of neural networks. However, most existing learning-based image compression approaches employ customized convolutional neural network (CNN) to utilize visual features by treating all pixels equally, neglecting the effect of local key features. Meanwhile, the convolutional filters in CNN usually express the local spatial relationship within the receptive field and seldom consider the long-range dependencies from distant locations. This results in the long-range dependencies of latent representations not being fully compressed. To address these issues, an end-to-end image compression method is proposed by integrating graph attention and asymmetric convolutional neural network (ACNN). Specifically, ACNN is used to strengthen the effect of local key features and reduce the cost of model training. Graph attention is introduced into image compression to address the bottleneck problem of CNN in modeling long-range dependencies. Meanwhile, regarding the limitation that existing attention mechanisms for image compression hardly share information, we propose a self-attention approach which allows information flow to achieve reasonable bit allocation. The proposed self-attention approach is in compliance with the perceptual characteristics of human visual system, as information can interact with each other via attention modules. Moreover, the proposed self-attention approach takes into account channel-level relationship and positional information to promote the compression effect of rich-texture regions. Experimental results demonstrate that the proposed method achieves state-of-the-art rate-distortion performances after being optimized by MS-SSIM compared to recent deep compression models on the benchmark datasets of Kodak and Tecnick.

关键词： Image coding Convolution Convolutional neural networks Training Kernel Visualization Rate-distortion Image compression graph attention network asymmetric convolutional neural network self-attention variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Joint Image and Feature Levels Disentanglement for Generalizable Vehicle Re-identification

引用

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 2023年第12期24卷 15259-15273页

作者： Kuang, Zhenyu He, Chuchu Huang, Yue Ding, Xinghao Li, Huafeng Xiamen Univ Sch Informat Fujian Key Lab Sensing & Comp Smart City Xiamen 361005 Peoples R China Kunming Univ Sci & Technol Yunnan Key Lab Artificial Intelligence Kunming 650500 Peoples R China

Domain generalization (DG), which doesn't require any data from target domains during training, is more challenging but practical than unsupervised domain adaptation (UDA). Since different vehicles of the same type have a similar appearance, neural networks always rely on a small amount of useful information to distinguish them, meaning that is more significant to remove ID-unrelated information for vehicle re-Identification (re-ID). Therefore, it is the key to eliminating the interference of a large amount of redundant information for the generalizable vehicle re-ID method. To address this unique challenge, we propose a novel disentanglement learning method that encourages variational autoencoder (VAE) network to reduce ID-unrelated features of vehicles by minimizing image reconstruction errors and providing sufficient representation to vehicle labels. To capture the intrinsic characteristics associated with the DG task, our core idea is to build the identity information streaming framework to separate ID-related and ID-unrelated information at the image and feature levels. In contrast with the general decoupling methods, our method leverages the decoupling of joint image and feature levels to extract more generalizable features. Furthermore, we present a brand-new vehicle dataset of truck types named "Optimus Prime (Opri)", which includes multiple images of each truck captured by cameras at different high-speed toll gates. Experimental results on public datasets demonstrate that our method can achieve promising results and outperform several state-of-the-art approaches. Our codes and models are available at JIFD.

关键词： Task analysis Feature extraction Image reconstruction Cameras Training Neural networks Interference Vehicle re-identification variational autoencoder domain generalization representation disentanglement

来源：评论

学校读者我要写书评

暂无评论

Pedestrian Trajectory Prediction Using Spatio-Temporal VAE 7th

Pedestrian Trajectory Prediction Using Spatio-Temporal VAE

引用

7th Chinese Conference on Pattern Recognition and Computer Vision

作者： Yu, Qing Xu, Zhenwei Zhou, Yaoyong Liu, Zhida Silamu, Wushouer Xinjiang Univ XJU Sch Comp Sci & Technol 777 Huarui St Urumqi 830017 Xinjiang Uygur Peoples R China Xinjiang Univ XJU Sch Software 499 Xibei Rd Urumqi 830091 Xinjiang Uygur Peoples R China

ISBN: (纸本)9789819785049;9789819785056

To address the high uncertainty of pedestrian behavior and the complexity of the environment at traffic intersections, this paper proposes an innovative Spatio-temporal variational autoencoder model (ST-VAE). This model aims to efficiently capture the spatio-temporal characteristics of pedestrian behavior to enhance the accuracy of individual trajectory prediction at traffic intersections. To effectively handle the complex relationships in pedestrian social interactions, a Social Graph Attention Network (SGAT) was designed. This network dynamically identifies the relative importance among traffic participants and optimizes interaction analysis. Furthermore, a Complex Gated Recurrent Unit (CGRU) was introduced to accurately capture the dynamic changes in time series data, thereby optimizing future trajectory prediction. In the final prediction results, a Final Position Clustering (FPC) method was designed as a post-processing technique to eliminate bias and increase prediction accuracy. Evaluation results on four public datasets showed that the ST-VAE model reduces the average absolute error (ADE) and the final displacement error (FDE) by 27.2% and 18.2% respectively compared to the best baseline model,SocialCVAE, thereby demonstrating its significant performance advantage. Further ablation studies revealed the specific contributions of each component to the model's performance, confirming the effectiveness and robustness of ST-VAE in handling diverse environments and complex interaction scenarios. The ST-VAE model not only provides a new perspective for pedestrian trajectory prediction technology but also offers solid technical support for the development of autonomous driving planning systems, showcasing the immense potential of deep learning in smart transportation system applications.

关键词： Pedestrian trajectory prediction Complex Gated Recurrent Unit variational autoencoder Social Graph Attention Network

来源：评论

学校读者我要写书评

暂无评论

A Method for Creating Realistic Synthetic Images Using a Generative Deep Learning Model for Classifying Anomalies in Panoramas

引用

PATTERN RECOGNITION AND IMAGE ANALYSIS 2024年第3期34卷 805-809页

作者： Arkhipov, P. O. Philippskih, S. L. Tsukanov, M. V. Russian Acad Sci Orel Branch Fed Res Ctr Comp Sci & Control Oryol 302025 Russia

This paper describes a method for generating realistic synthetic images using a generative deep learning model. When generating training examples, the features of climate, relief, architecture, and landscape of a particular area are considered. The generative model is represented by a variational autoencoder consisting of an encoder, a latent space, and a decoder. To train the autoencoder, a new dataset CSCDroneCL_v1 was created, consisting of manually labeled panoramic images and training examples selected from the open VisDrone2022 dataset. Based on the SOTA-ConvNet template, a discriminative neural network model was designed and trained to classify anomalies obtained from multitemporal panoramas. Architectural compatibility between generative and discriminative models is ensured by using a single design pattern. The unified architecture of neural network models made it possible to apply the transfer learning method. The use of pretrained parameters in the variational autoencoder model makes it possible to compensate for the small size of the CSCDroneCL_v1 dataset. The generated synthetic training examples of panoramas were added to the VisDrone2022 dataset. Using the new dataset, a discriminative model was trained, resulting in an increase in anomaly classification accuracy by 21.2% for the selected class.

关键词： panoramic image multiclass classification transfer learning variational autoencoder generative deep learning

来源：评论

学校读者我要写书评

暂无评论

Adaptive Gait Trajectory and Event Prediction of Lower Limb Exoskeletons for Various Terrains Using Reinforcement Learning

引用

JOURNAL OF INTELLIGENT & ROBOTIC SYSTEMS 2023年第2期109卷 1-20页

作者： Yu, Zhipeng Zhao, Jianghai Chen, Danhui Chen, Shuyan Wang, Xiaojie Chinese Acad Sci Inst Intelligent Machines Hefei Inst Phys Sci Hefei 230031 Peoples R China Univ Sci & Technol China Dept Sci Isl Hefei 230026 Peoples R China Univ Sci & Technol China Dept Automat Hefei 230026 Peoples R China Chinese Acad Sci Changzhou Inst Adv Mfg Technol Changzhou 213164 Peoples R China

The recent development of lower limb exoskeletons has focus on assisting walking on a variety of terrains, such as level ground, stairs, and ramps. However, smooth transitions between different locomotion modes are challenging due to unknown future information. In this paper, we propose a method that adaptively generates a complete gait cycle of trajectory and event prediction for lower limb exoskeletons during the transition stage. First, we design a generative model with the variational autoencoder (VAE) structure to parameterize the gait cycles of trajectory and events into low-dimensional latent representations. Then, the gait trajectory and event prediction are generated by VAE reconstruction, stride time and gait phase alignment. Lastly, reinforcement learning (RL) is adopted to minimize the process error of continuous gait trajectory and event prediction by finding the optimal latent representations of a generative model. The experiments collected gait data from passive and active exoskeletons. The results show that our method significantly outperforms the extend Kalman filter (EKF)-based method and moving horizon estimation (MHE)-based method for both gait trajectory and event prediction tasks during transitions between and within locomotion modes. The capability of our method gives the opportunity to enrich the control mode by providing future information and contributes a valuable tool for natural and smooth transitions among different terrains for the lower limb exoskeleton.

关键词： Lower limb exoskeleton Gait trajectory prediction Gait event prediction variational autoencoder Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

VAE-CoGAN: Unpaired image-to-image translation for low-level vision

引用

SIGNAL IMAGE AND VIDEO PROCESSING 2023年第4期17卷 1019-1026页

作者： Zhang, Juan Lang, Xiaoqi Huang, Bo Jiang, Xiaoyan Shanghai Univ Engn Sci Shanghai 201620 Peoples R China

Low-level vision problems, such as single image haze removal and single image rain removal, usually restore a clear image from an input image using a paired dataset. However, for many problems, the paired training dataset will not be available. In this paper, we propose an unpaired image-to-image translation method based on coupled generative adversarial networks (CoGAN) called VAE-CoGAN to solve this problem. Different from the basic CoGAN, we propose a shared-latent space and variational autoencoder (VAE) in framework. We use synthetic datasets and the real-world images to evaluate our method. The extensive evaluation and comparison results show that the proposed method can be effectively applied to numerous low-level vision tasks with favorable performance against the state-of-the-art methods.

关键词： Dehaze Derain Generative adversarial networks variational autoencoder

来源：评论

学校读者我要写书评

暂无评论

Inference over radiative transfer models using variational and expectation maximization methods

引用

MACHINE LEARNING 2023年第3期112卷 921-937页

作者： Svendsen, Daniel Heestermans Hernandez-Lobato, Daniel Martino, Luca Laparra, Valero Moreno-Martinez, Alvaro Camps-Valls, Gustau Univ Valencia Image Proc Lab IPL Valencia Spain Univ Autonoma Madrid Dept Comp Sci Madrid Spain Univ Rey Juan Carlos Madrid Dept Signal Theory & Telecommun Madrid Spain

Earth observation from satellites offers the possibility to monitor our planet with unprecedented accuracy. Radiative transfer models (RTMs) encode the energy transfer through the atmosphere, and are used to model and understand the Earth system, as well as to estimate the parameters that describe the status of the Earth from satellite observations by inverse modeling. However, performing inference over such simulators is a challenging problem. RTMs are nonlinear, non-differentiable and computationally costly codes, which adds a high level of difficulty in inference. In this paper, we introduce two computational techniques to infer not only point estimates of biophysical parameters but also their joint distribution. One of them is based on a variational autoencoder approach and the second one is based on a Monte Carlo Expectation Maximization (MCEM) scheme. We compare and discuss benefits and drawbacks of each approach. We also provide numerical comparisons in synthetic simulations and the real PROSAIL model, a popular RTM that combines land vegetation leaf and canopy modeling. We analyze the performance of the two approaches for modeling and inferring the distribution of three key biophysical parameters for quantifying the terrestrial biosphere.

关键词： variational autoencoder Expectation maximization Radiative transfer model Inverse modeling Density estimation

来源：评论

学校读者我要写书评

暂无评论

Improved history matching of channelized reservoirs using a novel deep learning-based parametrization method

GEOENERGY SCIENCE AND ENGINEERING

引用

GEOENERGY SCIENCE AND ENGINEERING 2023年 229卷

作者： Yousefzadeh, Reza Ahmadi, Mohammad Amirkabir Univ Technol Dept Petr Engn Tehran Iran

Most of the geological parametrization techniques used in history matching of sub-surface formations including the deep learning-based methods could not capture the non-linear and non-Gaussian dependencies and were limited to facies realizations. This paper proposes a novel method for parametrization of permeability realizations using ConvLSTM layers to build a variational autoencoder, named ConvLSTM-VAE, for the first time in this study. Since geological layers of the reservoir models are deposited during different periods, reservoir properties have a spatial distribution along the third dimension that can be regarded as a temporal distribution. To evaluate the proposed method, its parametrization results were compared with those of 3D convolutional VAE (3DCNN-VAE), which was recently used in the literature. Quantitative and visual results showed that the proposed method was far better than the 3DCNN-VAE, where the train, validation, and test losses obtained by the ConvLSTM-VAE was 0.3663, 0.3695, and 0.3710, and by the 3DCNN-VAE was 0.3980, 0.4086, 0.4095, respectively. In addition, history matching with parametrized realizations by the ConvLSTM-VAE could preserve the geologic realism in the conditioned realizations with significantly low data mismatch errors in the historical and prediction phases;however, history matching with 3DCNN-VAE parametrization and without any parametrization could not preserve the geologic realism. Finally, this paper can guide reservoir engineers to select a proper parametrization method for highly channelized, 3D, complex reservoirs with non-linear and non-Gaussian distribution of the uncertain parameters.

关键词： Data assimilation variational autoencoder History matching Deep learning Geological parametrization

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：