检索结果-内蒙古大学图书馆

DMMFnet: A Dual-Branch Multimodal Medical Image Fusion Network Using Super Token and Channel-Spatial Attention

INTERNATIONAL JOURNAL OF ADVANCED COMPUTER SCIENCE AND APPLICATIONS 2024年第8期15卷 696-705页

作者： Zhang, Yukun Wang, Lei Tahir, Muhammad Huang, Zizhen Han, Yaolong Yang, Shanliang Liu, Shilong Saeed, Muhammad Imran Shandong Univ Technol Sch Comp Sci & Technol Zibo 255000 Peoples R China Mohammad Ali Jinnah Univ Dept Comp Sci Block-6 Karachi 75400 Sindh Pakistan

-Multimodal medical image fusion leverages the correlation between different modal images to enhance the information contained within a single medical image. Existing fusion methods often fail to effectively extract multiscale features from medical images and establish long-distance relationships between deep feature blocks. To address these issues, we propose DMMFnet, an encoder-decoder fusion network that utilizes shared and private encoders to extract shared and private features. DMMFnet is based on super token sampling and channel-spatial attention. The shared encoder and decoder use a transformer structure with super token sampling technology to effectively integrate information from different modalities, improving processing efficiency and enhancing the ability to capture key features. The private encoder consists of invertible neural networks and transformer modules, designed to extract local and global features, respectively. A novel transformer module refines attention distribution and feature aggregation to capture superpixel-level global correlations, ensuring that the network effectively captures essential global information, thereby enhancing the quality of the fused image. Experimental results, comparing DMMFnet with nine leading fusion methods, indicate that DMMFnet significantly improves various evaluation metrics and achieves superior visual effects, demonstrating its advanced fusion capability.

关键词： Medical image fusion channel-spatial attention super token sampling encoder-decoder

来源：评论

学校读者我要写书评

暂无评论

An efficient transformer-based surrogate model with end-to-end training strategies for automatic history matching

GEOENERGY SCIENCE AND ENGINEERING

引用

GEOENERGY SCIENCE AND ENGINEERING 2024年 240卷

作者： Zhang, Jinding Kang, Jinzheng Zhang, Kai Zhang, Liming Liu, Piyang Liu, Xingyu Sun, Weijia Wang, Guangyao China Univ Petr East China Sch Petr Engn Qingdao Peoples R China Qingdao Univ Technol Sch Civil Engn Qingdao Peoples R China

In recent years, surrogate models have gained popularity as a tool to tackle the challenges presented by timeconsuming numerical simulations in automatic history matching (AHM). Although there are many different surrogate models designed to alleviate this issue, it is still challenging for most of them to handle the high dimensionality and strong non-linearity in reservoir models and dynamic production data. Inspired by the rapid development of Transformer, we propose a novel hybrid hierarchical Vision Transformer (HHVT) approach for history matching, which utilizes a unified architecture to predict production data for specific physical fields with an end-to-end strategy. For predicting the production data of wells, the Transformer supports parallelism computation among multiple timesteps, which shows more superiority than traditional recurrent neural networks. Specifically, our approach constructs a novel encoder-decoder Transformer architecture to learn the implicit features of high-level spatial parameters to match the features of time-series production data. With this architecture, HHVT achieves fast training and inference, which is suitable for large-scale datasets and highdimensional features. The proposed HHVT model is integrated with a multimodal optimization algorithm to find history-matching solutions. We first validated the effect of hyperparameters of HHVT on a simple 2D reservoir. Moreover, the proposed method was verified on the complex 3D Brugge model. The results demonstrate that the training speed of the Transformer-based model is approximately twice as fast as the surrogates based on convolutional and recurrent neural networks. The proposed HHVT also shows better prediction accuracy in two cases, compared with other surrogate models, which enhances the applicability of surrogate-based history-matching methods in large-scale complex reservoir scenarios.

关键词： History matching Surrogate model Transformer encoder-decoder Multimodal algorithm

来源：评论

学校读者我要写书评

暂无评论

Complex system anomaly detection via learnable temporal-spatial graph with degradation tendency segmentation

引用

ISA TRANSACTIONS 2024年 152卷 156-166页

作者： Han, Qinfeng Chen, Jinglong Wang, Jun Feng, Yong Xi An Jiao Tong Univ State Key Lab Mfg & Syst Engn Xian 710049 Peoples R China Xian Aerosp Prop Inst Sci & Technol Liquid Rocket Engine Lab Xian 710100 Peoples R China

To guarantee the safety and reliability of equipment operation, such as liquid rocket engine (LRE), carrying out system-level anomaly detection (AD) is crucial. However, current methods ignore the prior knowledge of mechanical system itself, and seldom unite the observations with the inherent relation in data tightly. Meanwhile, they neglect the weakness and nonindependence of system-level anomaly which is different from component fault. To overcome above limitations, we propose a separate reconstruction framework using worsened tendency for system-level AD. To prevent anomalous feature being attenuated, we first propose to divide single sample into two equal-length parts along the temporal dimension. And we maximize the mean maximum discrepancy (MMD) between feature segments to force encoders to learn normal features with different distributions. Then, to fully explore the multivariate time series, we model temporal-spatial dependence by temporal convolution and graph attention. Besides, a joint graph learning strategy is proposed to handle prior knowledge and data characteristics simultaneously. Finally, the proposed method is evaluated on two real multi-sensor datasets from LRE and the results demonstrate the effectiveness and potential of the proposed method on system-level AD.

关键词： Anomaly detection encoder-decoder Liquid rocket engine Multivariate time series Sample segmentation

来源：评论

学校读者我要写书评

暂无评论

Blind Source Separation of Electromagnetic Signals Based on Swish-Tasnet

引用

CIRCUITS SYSTEMS AND SIGNAL PROCESSING 2024年第10期43卷 6620-6636页

作者： Chen, Yang Liu, Jinming Mao, Jian Pang, Xiaoyu Jimei Univ Sch Comp Yinjiang Rd Xiamen 361021 Fujian Peoples R China

Digital devices may leak electromagnetic signals containing important information during operation, posing a risk of information leakage. In order to evaluate the safety level of the equipment, it is necessary to detect electromagnetic leakage signals and separate important information. Traditional methods such as ICA and IVA have limited performance when the signal input is smaller than the output. In recent years, signal separation technology based on deep learning has developed rapidly, demonstrating effective separation performance and potential in signal separation. For the separation of electromagnetic signals, we propose a blind source separation method for electromagnetic signals based on Swish-Tasnet. This method improves the time convolutional network structure and introduces the Swish activation function to optimize the internal functions and stacking layers of the model, thereby achieving efficient blind source separation of electromagnetic signals. The experiment evaluated the separation performance through SISDR, SDR, and SAR, and the results showed that this method performs excellently in reducing background noise interference and accurately separating signal sources from mixed signals. Swish-Tasnet not only provides a practical and effective technical means for electromagnetic signal leakage separation, but also brings substantial progress to the field of equipment information security assessment research. This method has broad prospects in practical applications and is worth further research and promotion.

关键词： Source separation Electromagnetic signals Deep learning encoder-decoder

来源：评论

学校读者我要写书评

暂无评论

Road extraction from remote sensing images based on a multi-scale asymmetric dual attention mechanism

引用

REMOTE SENSING LETTERS 2024年第8期15卷 751-761页

作者： Qu, Shenming Liu, Suchen Han, Fengyu Xie, Yuan Henan Univ Software Coll Henan Int Joint Lab Theories & Key Technol Intelli Kaifeng 475004 Henan Peoples R China

Aiming at the problems of road fracture and detail loss caused by not considering the geometric features in the road extraction method. We proposed an encoder-decoder architecture based on a multi-scale asymmetric dual attention mechanism. Firstly, A multi-scale convolution block in the shape of 'Union Jack' is designed. It includes symmetric convolution and asymmetric convolution along horizontal, vertical, left diagonal, and right diagonal spatial directions, and a multi-scale dilated convolution for extracting features of different scales. Remote dependence relationships are highly converged by using it, and road fracture problems caused by occlusion can be solved effectively. Secondly, a directional dual attention mechanism is proposed, which consists of directional channel attention using strip pooling and a directional spatial attention mechanism using asymmetric convolution along left diagonal and right diagonal spatial directions. It can use the directivity of asymmetric convolution to allocate attention mechanism adaptively in attention mechanism, and effectively avoid the road detail loss problem. Finally, we conducted corresponding experiments on the DeepGlobe and Ottawa road datasets, and the experimental results are superior to the current state-of-the-art methods.

关键词： Road extraction multi-scale asymmetric convolution asymmetric dual attention mechanism dilated convolution strip pooling encoder-decoder

来源：评论

学校读者我要写书评

暂无评论

Solving a 3D bin packing problem with stacking constraints

引用

COMPUTERS & INDUSTRIAL ENGINEERING 2024年 188卷

作者： Zhu, Weiwang Chen, Sheng Dai, Min Tao, Jiping Xiamen Univ Sch Aerosp Engn Xiamen Fujian Peoples R China Xiamen Univ Key Lab Big Data Intelligent Anal & Decis making F Xiamen Fujian Peoples R China

We consider a novel variant of the three-dimensional bin packing problem (3DBPP) arising from a maritime shipping port. Unlike the classical bin packing problem, our 3DBPP requires all items to be stacked vertically during the packing process. To tackle this problem effectively, we divide it into two subproblems: the stack packing problem (SPP) and the two-dimensional bin packing problem (2DBPP). We first formulate the SPP as an integer programming model to minimize the total bottom area of the stacks, which is further solved by a branch -and -price method. Then, in the 2DBPP, to pack all stacks into a minimum number of bins, we follow the left -down most extreme point placement strategy. An encoder-decoder model was developed to optimize the packing sequence and orientation of the stacks simultaneously. Industrial instances from shipping ports were collected and extended. The computational results show that our two -stage algorithm performs well in solving the stacking -constrained 3DBPP.

关键词： Three-dimensional bin packing problem Stacking constraints Branch-and-price encoder-decoder

来源：评论

学校读者我要写书评

暂无评论

Lower Limb Torque Prediction for Sit-To-Walk Strategies Using Long Short-Term Memory Neural Networks

引用

IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING 2024年 32卷 3977-3986页

作者： Perera, Chamalka Kenneth Gopalai, Alpha. A. Gouwanda, Darwin Ahmad, Siti. A. Teh, Pei-Lee Monash Univ Sch Engn Subang Jaya 47500 Selangor Malaysia Univ Putra Malaysia Fac Engn Serdang 43400 Malaysia Monash Univ Sch Business Subang Jaya 47500 Selangor Malaysia

Joint torque prediction is crucial when investigating biomechanics, evaluating treatments, and designing powered assistive devices. Controllers in assistive technology require reference torque trajectories to set the level of assistance for a patient during rehabilitation or when aiding essential daily tasks such as sit-to-walk (STW). STW itself can be generalized into strategies based on individual needs and movement patterns. In this study, three long short-term memory (LSTM) neural networks were empirically trained for hip and knee torque prediction considering these STW strategies and subject anthropometry. The hip and knee are the drivers of STW, while the network architectures were selected for recognizing temporal and spatial relationships. Performance of the LSTMs were compared and evaluated against the STW strategies to accurately generate strategy-specific and user-oriented torque. As such, train and test STW data were obtained from 65 subjects across three age groups: young, middle-aged, and older adults (19-73 years). Model inputs were hip and knee angles with horizontal center of mass velocity, while windowing allowed the LSTMs to dynamically adapt to real-time changes in STW transitions. The encoder-decoder LSTM showcased optimal performance with robust recognition of temporal features. It produced significantly (P <0.05) low hip and knee root mean square error (0.24 f 0.07 and 0.15 f 0.02 Nm/kg), strong Spearman's correlation (93.43 f 2.86 and 84.83 f 2.96%) and good intraclass correlation coefficients (greater than 0.75), demonstrating model reliability. Hence, this network predicts strategy and user oriented reference torques for personalized controllers in assistive devices, with more natural application of assistance.

关键词： Torque Hip Knee Mathematical models Long short term memory Biological system modeling Biomechanics Assistive devices Training Predictive models encoder-decoder CNN-LSTM strategy classification torque controllers assistive devices

来源：评论

学校读者我要写书评

暂无评论

Forwardformer: Efficient Transformer With Multi-Scale Forward Self-Attention for Day-Ahead Load Forecasting

引用

IEEE TRANSACTIONS ON POWER SYSTEMS 2024年第1期39卷 1421-1433页

作者： Qu, Kai Si, Gangquan Shan, Zihan Wang, Qianyue Liu, Xiaoliang Yang, Chen Xi An Jiao Tong Univ Sch Elect Engn State Key Lab Elect Insulat & Power Equipment Xian 710049 Peoples R China

Accurate load forecasting can maintain the safety and stability of power grids. The mainstream models are based on complex recurrent or convolutional neural networks (RNNs, CNNs), and in recent years they are often used in combination with the attention mechanism. The shortcomings of these models are that they cannot get rid of sequential computation and fail to capture long-term dependence. For better application in day-ahead load forecasting (DALF), we propose a new network architecture, i.e., Forwardformer, which is implemented by imposing some effective improvements on the Transformer (a pioneering network model with applications in Natural Language Processing (NLP)). The core of Forwardformer is the multi-scale forward self-attention (MSFSA) and the correction structure of the encoder-dual decoder, which confer better computational efficiency and forecasting accuracy. Meanwhile, to improve forecasting accuracy on special days (weekends, holidays, etc.), the MSFSA configures dilated attention and global attention for them, respectively. Experiments performed on datasets from China and America demonstrated that the Forwardformer requires less runtime while being superior in forecasting accuracy. Especially in terms of weekends and holidays, it has outstanding advantages and provides a new solution to the DALF problem.

关键词： Load modeling Forecasting Load forecasting Predictive models Transformers Time series analysis Power system stability Correction structure encoder-decoder forwardformer holidays multi-scale forward self-attention short-term load forecasting weekends

来源：评论

学校读者我要写书评

暂无评论

A Spatial-Temporal Graph Model for Pronunciation Feature Prediction of Chinese Poetry

引用

IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 2023年第12期34卷 10294-10308页

作者： Wang, Qing Liu, Weiping Wang, Xiumei Chen, Xinghong Chen, Guannan Wu, Qingxiang Fujian Normal Univ Minist Educ Key Lab Optoelect Sci & Technol Med Fuzhou 350007 Peoples R China Fujian Normal Univ Fujian Prov Key Lab Photon Technol Fuzhou 350007 Peoples R China

With the development of artificial intelligence, speech recognition and prediction have become one of the important research domains with wild applications, such as intelligent control, education, individual identification, and emotion analysis. Chinese poetry reading contains rich features of continuous pronunciations, such as mood, emotion, rhythm schemes, lyric reading, and artistic expression. Therefore, the prediction of the pronunciation characteristics of a Chinese poetry reading is the significance for the presentation of high-level machine intelligence and has the potential to create a high-level intelligent system for teaching children to read Tang poetry. Mel frequency cepstral coefficient (MFCC) is currently used to present important speech features. Due to the complexity and high degree of nonlinearity in poetry reading, however, there is a tough challenge facing accurate pronunciation feature prediction, that is, how to model complex spatial correlations and time dynamics, such as rhyme schemes. As for many current methods, they ignore the spatial and temporal characteristics in MFCC presentation. In addition, these methods are subjected to certain limitations on prediction for long-term performance. In order to solve these problems, we propose a novel spatial-temporal graph model (STGM-MHA) based on multihead attention for the purpose of pronunciation feature prediction of Chinese poetry. The STGM-MHA is designed using an encoder-decoder structure. The encoder compresses the data into a hidden space representation, while the decoder reconstructs the hidden space representation as output. In the model, a novel gated recurrent unit (GRU) module (AGRU) based on multihead attention is proposed to extract the spatial and temporal features of MFCC data effectively. The evaluation comparison of our proposed model versus state-of-the-art methods in six datasets reveals the clear advantage of the proposed model.

关键词： Mel frequency cepstral coefficient Predictive models Rhythm Feature extraction Analytical models Data models Computational modeling AGRU Chinese poetry encoder-decoder graph modeling Mel frequency cepstral coefficient (MFCC) pronunciation features spatial-temporal graph model (STGM-MHA)

来源：评论

学校读者我要写书评

暂无评论

FasterMDE: A real-time monocular depth estimation search method that balances accuracy and speed on the edge

引用

APPLIED INTELLIGENCE 2023年第20期53卷 24566-24586页

作者： ZiWen, Dou YuQi, Li Dong, Ye Harbin Inst Technol Sch Instrumentat Sci & Engn 92 Xidazhi St Harbin 150001 Peoples R China

Monocular depth estimation (MDE) is critical in enabling intelligent autonomous systems and has received considerable attention in recent years. Achieving both low latency and high accuracy in MDE is desirable but challenging to optimize, especially on edge devices. In this paper, we present a novel approach to balancing speed and accuracy in MDE on edge devices. We introduce FasterMDE, an efficient and fast encoder-decoder network architecture that leverages a multiobjective neural architecture search method to find the optimal encoder structure for the target edge. Moreover, we incorporate a neural window fully connected CRF module into the network as the decoder, enhancing fine-grained depth prediction based on coarse depth and image features. To address the issue of bad "local minimums" in the multiobjective neural architecture search, we propose a new approach for automatically learning the weights of subobjective loss functions based on uncertainty. We also accelerate the FasterMDE model using TensorRT and implement it on a target edge device. The experimental results demonstrate that FasterMDE achieves a better balance of speed and accuracy on the KITTI and NYUv2 datasets compared to previous methods. We validate the effectiveness of the proposed method through an ablation study and verify the real-time monocular depth estimation performance of FasterMDE in realistic scenarios. On the KITTI dataset, the FasterMDE model achieves a high frame rate of 555.55 FPS with 9.1% Abs Rel on a single NVIDIA Titan RTX GPU and 14.46 FPS on the NVIDIA Jetson Xavier NX.

关键词： Monocular depth estimation Real-time Edge computing Multi-objective neural architecture search encoder-decoder

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：