检索结果-南通市图书馆

Diabetic Retinopathy Classification Based on Bilinear Cross Attention network

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

Journal of Computer and Communications 2023年第5期11卷 16-28页

作者： Zhiyuan Ren Chen Xing School of Computer Science and Technology Qilu University of Technology (Shandong Academy of Sciences) Jinan China

Computer-aided diagnostic systems can assist doctors in diagnosing and treating DR cases more effectively, thereby improving work efficiency, reducing the burden on doctors during examinations, and alleviating problems related to uneven distribution of medical resources and shortage of doctors. In this article, we propose a classification method for diabetic retinopathy based on a bilinear multi-attention network. This method uses two backbone networks to extract features, and cross-shares the features using two attention modules to further deepen feature extraction. The non-local attention module is added to address the limitations of traditional convolutional neural networks in capturing global information. By paying attention to highly correlated pathological areas globally, performance improvement can be achieved. We achieved an accuracy of 91.7% on the Messidor dataset.

关键词： Diabetic Retinopathy deep neural network Styling deep neural network

FastAttacker: Semantic Perturbation Functions via Three Classifications

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Journal of Information Security 2023年第2期14卷 181-194页

作者： Meng Lu Institute of Automation Chinese Academy of Sciences Beijing China

deep neural networks (DNNs) have achieved great success in tasks such as image classification, speech recognition, and natural language processing. However, they are susceptible to false predictions caused by adversarial exemplars, which are normal inputs with imperceptible perturbations. Adversarial samples have been widely studied in image classification, but not as much in text classification. Current textual attack methods often rely on low-success-rate heuristic replacement strategies at the character or word level, which cannot search for the best solution while maintaining semantic consistency and linguistic fluency. Our framework, FastAttacker, generates natural adversarial text efficiently and effectively by constructing different semantic perturbation functions. It optimizes perturbations constrained in generic semantic spaces, such as the typo space, knowledge space, contextualized semantic space, or a combination. As a result, the generated adversarial texts are semantically close to the original inputs. Experiments show that FastAttacker generates adversarial texts from different levels of spatial constraints, making the problem of finding synonyms an optimal solution problem. Our approach is not only robust in terms of attack generation, but also in terms of adversarial defense. Experiments have shown that state-of-the-art language models and defense strategies are still vulnerable to FastAttack attacks.

关键词： FastAttack Text Learning deep neural network

An Image Segmentation Algorithm Based on a Local Region Conditional Random Field Model

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

International Journal of Communications, network and System Sciences 2020年第9期13卷 139-159页

作者： Xiao Jiang Haibin Yu Shuaishuai Lv School of Mechanical Engineering Hangzhou Dianzi University Hangzhou China School of Electronics and Information Hangzhou Dianzi University Hangzhou China

To reduce the computation cost of a combined probabilistic graphical model and a deep neural network in semantic segmentation, the local region condition random field (LRCRF) model is investigated which selectively applies the condition random field (CRF) to the most active region in the image. The full convolutional network structure is optimized with the ResNet-18 structure and dilated convolution to expand the receptive field. The tracking networks are also improved based on SiameseFC by considering the frame relations in consecutive-frame traffic scene maps. Moreover, the segmentation results of the greyscale input data sets are more stable and effective than using the RGB images for deep neural network feature extraction. The experimental results show that the proposed method takes advantage of the image features directly and achieves good real-time performance and high segmentation accuracy.

关键词： Image Segmentation Local Region Condition Random Field Model deep neural network Consecutive Shooting Traffic Scene

Forecast electricity demand in commercial building with machine learning models to enable demand response programs

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

能源与人工智能(英文) 2022年第1期7卷 18-30页

作者： Fabiano Pallonetto Changhong Jin Eleni Mangina School of Computer Science University College DublinIreland School of Business Maynooth UniversityIreland

Electricity load forecasting is an important part of power system *** forecasting electricity load have great impact on a number of departments in power *** to electricity load simulation(white-box model),electricity load forecasting(black-box model)does not require expertise in building *** development cycle of the electricity load forecasting model is much shorter than the design cycle of the electricity load *** developments in machine learning have lead to the creation of models with strong fitting and accuracy to deal with nonlinear *** on the real load dataset,this paper evaluates and compares the two mainstream short-term load forecasting *** the experiment,this paper first enumerates the common methods of short-term load forecasting and explains the principles of Long Short-term Memory networks(LSTMs)and Support Vector Machines(SVM)used in this ***,based on the characteristics of the electricity load dataset,data pre-processing and feature selection takes *** paper describes the results of a controlled experiment to study the importance of feature *** LSTMs model and SVM model are applied to one-hour ahead load forecasting and one-day ahead peak and valley load *** predictive accuracy of these models are calculated based on the error between the actual and predicted loads,and the runtime of the model is *** results show that the LSTMs model have a higher prediction accuracy when the load data is ***,the overall performance of the SVM model is better when the load data used to train the model is insufficient and the time cost is prioritized.

关键词： deep neural network Model assessment Short-term load forecasting Feature selection Support Vector Machines Artificial neural networks Long Short-term Memory networks Demand response

Towards online optimisation of solid oxide fuel cell performance: Combining deep learning with multi-physics simulation

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

能源与人工智能(英文) 2020年第1期1卷 11-21页

作者： Haoran Xu Jingbo Ma Peng Tan Bin Chen Zhen Wu Yanxiang Zhang Huizhi Wang Jin Xuan Meng Ni Department of Chemical Engineering Loughborough UniversityLoughboroughUnited Kingdom National Key Laboratory for Precision Hot Processing of Metals School of Materials Science and EngineeringHarbin Institute of TechnologyHarbin 150001HeilongjiangChina Department of Thermal Science and Energy Engineering University of Science and Technology of ChinaHefei 230026China Institute of deep Earth Sciences and Green Energy Shenzhen UniversityShenzhen 518060China Shaanxi Key Laboratory of Energy Chemical Process Intensification School of Chemical Engineering and TechnologyXi’an Jiaotong UniversityXi’anChina Department of Mechanical Engineering Imperial College LondonLondon SW72AZ United Kingdom Department of Building and Real Estate Building Energy Research GroupThe Hong Kong Polytechnic UniversityHung HomKowloonHong KongChina Environmental Energy Research Group Research Institute for Sustainable Urban Development(RISUD)The Hong Kong Polytechnic UniversityHung HomKowloonHong KongChina

The use of solid oxide fuel cells(SOFCs)is a promising approach towards achieving sustainable electricity pro-duction from *** utilisation of the hydrocarbons and biomass in SOFCs is particularly attractive owing to their wide distribution,high energy density,and low *** long-term operation of SOFCs using such fuels remains difficult owing to a lack of an effective diagnosis and optimisation system,which requires not only a precise analysis but also a fast *** this study,we developed a hybrid model for an on-line analysis of SOFCs at the cell *** model combines a multi-physics simulation(MPS)and deep learning,overcoming the complexity of MPS for a model-based control system,and reducing the cost of building a database(compared with the experiments)for the training of a deep neural *** maximum temperature gradient and heat generation are two target parameters for an efficient operation of *** results show that a precise predic-tion can be achieved from a trained AI algorithm,in which the relative error between the MPS and AI models is less than 1%.Moreover,an online optimisation is realised using a genetic algorithm,achieving the maximum power density within the limitations of the temperature gradient and operating *** method can also be applied to the prediction and optimisation of other non-liner,dynamic systems.

关键词： Solid oxide fuel cell Multi-physics simulation Artificial intelligence deep neural network Hybrid model On-line optimisation

A Lightweight Temporal Convolutional network for Human Motion Prediction

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Transactions of Nanjing University of Aeronautics and Astronautics 2022年第S1期39卷 150-157页

作者： WANG You QIAO Bing College of Astronautics Nanjing University of Aeronautics and AstronauticsNanjing 210016P.R.China

A lightweight multi-layer residual temporal convolutional network model(RTCN)is proposed to target the highly complex kinematics and temporal correlation of human *** uses 1-D convolution to efficiently obtain the spatial structure information of human motion and extract the correlation in the time series of human *** residual structure is applied to the proposed network model to alleviate the problem of gradient disappearance in the deep *** on the Human 3.6M dataset demonstrate that the proposed method effectively reduces the errors of motion prediction compared with previous methods,especially of long-term prediction.

关键词： human motion prediction temporal convolutional network short-term prediction long-term prediction deep neural network

Bregman Learning for Generative Adversarial networks

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

Bregman Learning for Generative Adversarial Networks

Koopman-Based Optimal Tracking Control of Nonlinear Dynamical Systems

第30届中国控制与决策会议

作者： Jian Gao Hamidou Tembine Learning ＆ Game Theory Laboratory Engineering DivisionNew York University

In this paper we develop a game theoretic learning framework for deep generative ***,the problem of minimizing the dissimilarity between the generator distribution and real data is introduced based on ***,the optimization problem is transformed into a zero-sum game with two adversarial players,and the existence of Nash equilibrium is established in the quasi-concave-convex case under suitable ***,a general Bregman-based learning algorithm is proposed to find the Nash *** algorithm is proved to have a doubly logarithmic convergence time with respect to the precision of the minimax value in potential convex ***,our methodology is implemented in three application scenarios and compared with several existing optimization *** qualitative and quantitative evaluation show that the generative model trained by our algorithm has the state-of-art performance.

关键词： Bregman learning GAN Convex Optimization deep neural network Game Theory

来源： cnki会议评论

在线全文

cnki会议

学校读者我要写书评

暂无评论

Koopman-Based Optimal Tracking Control of Nonlinear Dynamica...

A real-time ECG signal classification algorithm

第40届中国控制会议

作者： Xuyang Lou Ling Ma Qian Ye School of IoT Engineering Jiangnan University Wuxi Institute of Technology

A data-driven procedure to approximate the Koopman operator is a leading technology in researching the properties of nonlinear dynamical systems. The Koopman operator is an infinite-dimensional linear operator that can be used to predict and control a nonlinear system in a linear manner. In this paper, the reformulation and control problems of complex nonlinear dynamical systems are explored. The extended dynamical mode decomposition method and the deep neural network method are applied, respectively, to obtain a finite approximation of the Koopman operator. Then, we present Koopman-based optimal tracking controller for nonlinear dynamical systems by using the linear quadratic tracking control strategy. Finally, simulation results demonstrate that the linear embedding of nonlinear systems by data-driven procedure can approximate the states of the nonlinear systems and the desired output of nonlinear systems can be optimally tracked based on the designed controller of the linear embedding system.

关键词： Koopman operator data-driven control dynamical mode decomposition deep neural network linear quadratic tracking control

来源： cnki会议评论

在线全文

cnki会议

学校读者我要写书评

暂无评论

A real-time ECG signal classification algorithm