Predicting disruptions across different tokamaks is necessary for next generation device.Future large-scale tokamaks can hardly tolerate disruptions at high performance discharge,which makes it difficult for current d...Predicting disruptions across different tokamaks is necessary for next generation device.Future large-scale tokamaks can hardly tolerate disruptions at high performance discharge,which makes it difficult for current data-driven methods to obtain an acceptable result.A machine learning method capable of transferring a disruption prediction model trained on one tokamak to another is required to solve the problem.The key is a feature extractor which is able to extract common disruption precursor traces in tokamak diagnostic data,and can be easily transferred to other tokamaks.Based on the concerns above,this paper presents a deep feature extractor,namely,the fusion feature extractor(FFE),which is designed specifically for extracting disruption precursor features from common diagnostics on tokamaks.Furthermore,an FFE-based disruption predictor on J-TEXT is demonstrated.The feature extractor is aimed to extracting disruption-related precursors and is designed according to the precursors of disruption and their representations in common tokamak diagnostics.Strong inductive bias on tokamak diagnostics data is introduced.The paper presents the evolution of the neural network feature extractor and its comparison against general deep neural networks,as well as a physics-based feature extraction with a traditional machine learning method.Results demonstrate that the FFE may reach a similar effect with physics-guided manual feature extraction,and obtain a better result compared with other deep learning methods.展开更多
Accessible communication based on sign language recognition(SLR)is the key to emergency medical assistance for the hearing-impaired community.Balancing the capture of both local and global information in SLR for emerg...Accessible communication based on sign language recognition(SLR)is the key to emergency medical assistance for the hearing-impaired community.Balancing the capture of both local and global information in SLR for emergency medicine poses a significant challenge.To address this,we propose a novel approach based on the inter-learning of visual features between global and local information.Specifically,our method enhances the perception capabilities of the visual feature extractor by strategically leveraging the strengths of convolutional neural network(CNN),which are adept at capturing local features,and visual transformers which perform well at perceiving global features.Furthermore,to mitigate the issue of overfitting caused by the limited availability of sign language data for emergency medical applications,we introduce an enhanced short temporal module for data augmentation through additional subsequences.Experimental results on three publicly available sign language datasets demonstrate the efficacy of the proposed approach.展开更多
The application of transformer networks and feature fusion models in medical image segmentation has aroused considerable attention within the academic circle.Nevertheless,two main obstacles persist:(1)the restrictions...The application of transformer networks and feature fusion models in medical image segmentation has aroused considerable attention within the academic circle.Nevertheless,two main obstacles persist:(1)the restrictions of the Transformer network in dealing with locally detailed features,and(2)the considerable loss of feature information in current feature fusion modules.To solve these issues,this study initially presents a refined feature extraction approach,employing a double-branch feature extraction network to capture complex multi-scale local and global information from images.Subsequently,we proposed a low-loss feature fusion method-Multi-branch Feature Fusion Enhancement Module(MFFEM),which realizes effective feature fusion with minimal loss.Simultaneously,the cross-layer cross-attention fusion module(CLCA)is adopted to further achieve adequate feature fusion by enhancing the interaction between encoders and decoders of various scales.Finally,the feasibility of our method was verified using the Synapse and ACDC datasets,demonstrating its competitiveness.The average DSC(%)was 83.62 and 91.99 respectively,and the average HD95(mm)was reduced to 19.55 and 1.15 respectively.展开更多
Deep learning has shown its human-level performance in various applications.However,current deep learning models are characterized by catastrophic forgetting of old knowledge when learning new classes.This poses a cha...Deep learning has shown its human-level performance in various applications.However,current deep learning models are characterized by catastrophic forgetting of old knowledge when learning new classes.This poses a challenge such as in intelligent diagnosis systems where initially only training data of a limited number of diseases are available.In this case,updating the intelligent system with data of new diseases would inevitably downgrade its performance on previously learned diseases.Inspired by the process of learning new knowledge in human brains,we propose a Bayesian generative model for continual learning built on afixed pre-trained feature extractor.In this model,knowledge of each old class can be compactly represented by a collection of statistical distributions,e.g.,with Gaussian mixture models,and naturally kept from forgetting in continual learning over time.Unlike existing class-incremental learning methods,the proposed approach is not sensitive to the continual learning process and can be additionally well applied to the data-incremental learning scenario.Experiments on multiple medical and natural image classification tasks reveal that the proposed approach outperforms state-of-the-art approaches that even keep some images of old classes during continual learning of new classes.展开更多
Research on human motion prediction has made significant progress due to its importance in the development of various artificial intelligence applications.However,effectively capturing spatio-temporal features for smo...Research on human motion prediction has made significant progress due to its importance in the development of various artificial intelligence applications.However,effectively capturing spatio-temporal features for smoother and more precise human motion prediction remains a challenge.To address these issues,a robust human motion prediction method via integration of spatial and temporal cues(RISTC)has been proposed.This method captures sufficient spatio-temporal correlation of the observable sequence of human poses by utilizing the spatio-temporal mixed feature extractor(MFE).In multi-layer MFEs,the channel-graph united attention blocks extract the augmented spatial features of the human poses in the channel and spatial dimension.Additionally,multi-scale temporal blocks have been designed to effectively capture complicated and highly dynamic temporal information.Our experiments on the Human3.6M and Carnegie Mellon University motion capture(CMU Mocap)datasets show that the proposed network yields higher prediction accuracy than the state-of-the-art methods.展开更多
With the increasing complexity of industrial processes, the high-dimensional industrial data exhibit a strong nonlinearity, bringing considerable challenges to the fault diagnosis of industrial processes. To efficient...With the increasing complexity of industrial processes, the high-dimensional industrial data exhibit a strong nonlinearity, bringing considerable challenges to the fault diagnosis of industrial processes. To efficiently extract deep meaningful features that are crucial for fault diagnosis, a sparse Gaussian feature extractor(SGFE) is designed to learn a nonlinear mapping that projects the raw data into the feature space with the fault label dimension. The feature space is described by the one-hot encoding of the fault category label as an orthogonal basis. In this way, the deep sparse Gaussian features related to fault categories can be gradually learned from the raw data by SGFE. In the feature space,the sparse Gaussian(SG) loss function is designed to constrain the distribution of features to multiple sparse multivariate Gaussian distributions. The sparse Gaussian features are linearly separable in the feature space, which is conducive to improving the accuracy of the downstream fault classification task. The feasibility and practical utility of the proposed SGFE are verified by the handwritten digits MNIST benchmark and Tennessee-Eastman(TE) benchmark process,respectively.展开更多
In industrial process control systems,there is overwhelming evidence corroborating the notion that economic or technical limitations result in some key variables that are very difficult to measure online.The data-driv...In industrial process control systems,there is overwhelming evidence corroborating the notion that economic or technical limitations result in some key variables that are very difficult to measure online.The data-driven soft sensor is an effective solution because it provides a reliable and stable online estimation of such variables.This paper employs a deep neural network with multiscale feature extraction layers to build soft sensors,which are applied to the benchmarked Tennessee-Eastman process(TEP)and a real wind farm case.The comparison of modelling results demonstrates that the multiscale feature extraction layers have the following advantages over other methods.First,the multiscale feature extraction layers significantly reduce the number of parameters compared to the other deep neural networks.Second,the multiscale feature extraction layers can powerfully extract dataset characteristics.Finally,the multiscale feature extraction layers with fully considered historical measurements can contain richer useful information and improved representation compared to traditional data-driven models.展开更多
Poultry feeding behaviors provide valuable information for system design and farm management.This study developed poultry feeding behavior detectors using the faster region-based convolution neural network(faster R-CN...Poultry feeding behaviors provide valuable information for system design and farm management.This study developed poultry feeding behavior detectors using the faster region-based convolution neural network(faster R-CNN).Twenty 50-day-old Jingfen layer pullets were kept in four experimental compartments and could freely move between adjacent ones.Four light colors(white,red,green,and blue)were supplied to create environmental variations for detector development.A camera was installed atop each compartment to capture images for detector development.Several hyperparameters were finetuned to determine the optimal one.Based on the trade-off strategies between detection accuracy and processing speed,the following strategies were deployed to develop the detector:feature extractor of inception V2,the model trained with common objects in context dataset,fixed_shape_resizer with the size of 600×600 pixels,kernel stride of 8300 proposals,and dynamic learning rate.The final detector had 95.7%recall,94.2%average precision,94.9%F1 score,23.5 mm root mean square error,and 8.3 fps processing speed,indicating decent performance for detecting poultry feeding behaviors.With the trained detector,temporal and spatial feeding behaviors of individual poultry can be successfully characterized.It is concluded that the faster RCNN should be a useful tool to continuously monitor poultry feeding behaviors in group settings.展开更多
基金Project supported by the National Key R&D Program of China (Grant No. 2022YFE03040004)the National Natural Science Foundation of China (Grant No. 51821005)
文摘Predicting disruptions across different tokamaks is necessary for next generation device.Future large-scale tokamaks can hardly tolerate disruptions at high performance discharge,which makes it difficult for current data-driven methods to obtain an acceptable result.A machine learning method capable of transferring a disruption prediction model trained on one tokamak to another is required to solve the problem.The key is a feature extractor which is able to extract common disruption precursor traces in tokamak diagnostic data,and can be easily transferred to other tokamaks.Based on the concerns above,this paper presents a deep feature extractor,namely,the fusion feature extractor(FFE),which is designed specifically for extracting disruption precursor features from common diagnostics on tokamaks.Furthermore,an FFE-based disruption predictor on J-TEXT is demonstrated.The feature extractor is aimed to extracting disruption-related precursors and is designed according to the precursors of disruption and their representations in common tokamak diagnostics.Strong inductive bias on tokamak diagnostics data is introduced.The paper presents the evolution of the neural network feature extractor and its comparison against general deep neural networks,as well as a physics-based feature extraction with a traditional machine learning method.Results demonstrate that the FFE may reach a similar effect with physics-guided manual feature extraction,and obtain a better result compared with other deep learning methods.
基金supported by the National Natural Science Foundation of China(No.62376197)the Tianjin Science and Technology Program(No.23JCYBJC00360)the Tianjin Health Research Project(No.TJWJ2025MS045).
文摘Accessible communication based on sign language recognition(SLR)is the key to emergency medical assistance for the hearing-impaired community.Balancing the capture of both local and global information in SLR for emergency medicine poses a significant challenge.To address this,we propose a novel approach based on the inter-learning of visual features between global and local information.Specifically,our method enhances the perception capabilities of the visual feature extractor by strategically leveraging the strengths of convolutional neural network(CNN),which are adept at capturing local features,and visual transformers which perform well at perceiving global features.Furthermore,to mitigate the issue of overfitting caused by the limited availability of sign language data for emergency medical applications,we introduce an enhanced short temporal module for data augmentation through additional subsequences.Experimental results on three publicly available sign language datasets demonstrate the efficacy of the proposed approach.
基金funded by the Henan Science and Technology research project(222103810042)Support by the open project of scientific research platform of grain information processing center of Henan University of Technology(KFJJ-2021-108)+1 种基金Support by the innovative funds plan of Henan University of Technology(2021ZKCJ14)Henan University of Technology Youth Backbone Teacher Program.
文摘The application of transformer networks and feature fusion models in medical image segmentation has aroused considerable attention within the academic circle.Nevertheless,two main obstacles persist:(1)the restrictions of the Transformer network in dealing with locally detailed features,and(2)the considerable loss of feature information in current feature fusion modules.To solve these issues,this study initially presents a refined feature extraction approach,employing a double-branch feature extraction network to capture complex multi-scale local and global information from images.Subsequently,we proposed a low-loss feature fusion method-Multi-branch Feature Fusion Enhancement Module(MFFEM),which realizes effective feature fusion with minimal loss.Simultaneously,the cross-layer cross-attention fusion module(CLCA)is adopted to further achieve adequate feature fusion by enhancing the interaction between encoders and decoders of various scales.Finally,the feasibility of our method was verified using the Synapse and ACDC datasets,demonstrating its competitiveness.The average DSC(%)was 83.62 and 91.99 respectively,and the average HD95(mm)was reduced to 19.55 and 1.15 respectively.
基金supported in part by the National Natural Science Foundation of China(Grant Nos.62071502,U1811461)the Guangdong Key Research and Development Program(Grant No.2020B1111190001).
文摘Deep learning has shown its human-level performance in various applications.However,current deep learning models are characterized by catastrophic forgetting of old knowledge when learning new classes.This poses a challenge such as in intelligent diagnosis systems where initially only training data of a limited number of diseases are available.In this case,updating the intelligent system with data of new diseases would inevitably downgrade its performance on previously learned diseases.Inspired by the process of learning new knowledge in human brains,we propose a Bayesian generative model for continual learning built on afixed pre-trained feature extractor.In this model,knowledge of each old class can be compactly represented by a collection of statistical distributions,e.g.,with Gaussian mixture models,and naturally kept from forgetting in continual learning over time.Unlike existing class-incremental learning methods,the proposed approach is not sensitive to the continual learning process and can be additionally well applied to the data-incremental learning scenario.Experiments on multiple medical and natural image classification tasks reveal that the proposed approach outperforms state-of-the-art approaches that even keep some images of old classes during continual learning of new classes.
基金supported by the National Key R&D Program of China(No.2018YFB1305200)the Natural Science Foundation of Zhejiang Province(No.LGG21F030011)。
文摘Research on human motion prediction has made significant progress due to its importance in the development of various artificial intelligence applications.However,effectively capturing spatio-temporal features for smoother and more precise human motion prediction remains a challenge.To address these issues,a robust human motion prediction method via integration of spatial and temporal cues(RISTC)has been proposed.This method captures sufficient spatio-temporal correlation of the observable sequence of human poses by utilizing the spatio-temporal mixed feature extractor(MFE).In multi-layer MFEs,the channel-graph united attention blocks extract the augmented spatial features of the human poses in the channel and spatial dimension.Additionally,multi-scale temporal blocks have been designed to effectively capture complicated and highly dynamic temporal information.Our experiments on the Human3.6M and Carnegie Mellon University motion capture(CMU Mocap)datasets show that the proposed network yields higher prediction accuracy than the state-of-the-art methods.
基金Projects(62125306, 62133003) supported by the National Natural Science Foundation of ChinaProject(TPL2019C03) supported by the Open Fund of Science and Technology on Thermal Energy and Power Laboratory,ChinaProject supported by the Fundamental Research Funds for the Central Universities(Zhejiang University NGICS Platform),China。
文摘With the increasing complexity of industrial processes, the high-dimensional industrial data exhibit a strong nonlinearity, bringing considerable challenges to the fault diagnosis of industrial processes. To efficiently extract deep meaningful features that are crucial for fault diagnosis, a sparse Gaussian feature extractor(SGFE) is designed to learn a nonlinear mapping that projects the raw data into the feature space with the fault label dimension. The feature space is described by the one-hot encoding of the fault category label as an orthogonal basis. In this way, the deep sparse Gaussian features related to fault categories can be gradually learned from the raw data by SGFE. In the feature space,the sparse Gaussian(SG) loss function is designed to constrain the distribution of features to multiple sparse multivariate Gaussian distributions. The sparse Gaussian features are linearly separable in the feature space, which is conducive to improving the accuracy of the downstream fault classification task. The feasibility and practical utility of the proposed SGFE are verified by the handwritten digits MNIST benchmark and Tennessee-Eastman(TE) benchmark process,respectively.
基金supported by National Natural Science Foundation of China(No.61873142)the Science and Technology Research Program of the Chongqing Municipal Education Commission,China(Nos.KJZD-K202201901,KJQN202201109,KJQN202101904,KJQN202001903 and CXQT21035)+2 种基金the Scientific Research Foundation of Chongqing University of Technology,China(No.2019ZD76)the Scientific Research Foundation of Chongqing Institute of Engineering,China(No.2020xzky05)the Chongqing Municipal Natural Science Foundation,China(No.cstc2020jcyj-msxmX0666).
文摘In industrial process control systems,there is overwhelming evidence corroborating the notion that economic or technical limitations result in some key variables that are very difficult to measure online.The data-driven soft sensor is an effective solution because it provides a reliable and stable online estimation of such variables.This paper employs a deep neural network with multiscale feature extraction layers to build soft sensors,which are applied to the benchmarked Tennessee-Eastman process(TEP)and a real wind farm case.The comparison of modelling results demonstrates that the multiscale feature extraction layers have the following advantages over other methods.First,the multiscale feature extraction layers significantly reduce the number of parameters compared to the other deep neural networks.Second,the multiscale feature extraction layers can powerfully extract dataset characteristics.Finally,the multiscale feature extraction layers with fully considered historical measurements can contain richer useful information and improved representation compared to traditional data-driven models.
基金funded by China Scholar Council(Grant No.201808410600).
文摘Poultry feeding behaviors provide valuable information for system design and farm management.This study developed poultry feeding behavior detectors using the faster region-based convolution neural network(faster R-CNN).Twenty 50-day-old Jingfen layer pullets were kept in four experimental compartments and could freely move between adjacent ones.Four light colors(white,red,green,and blue)were supplied to create environmental variations for detector development.A camera was installed atop each compartment to capture images for detector development.Several hyperparameters were finetuned to determine the optimal one.Based on the trade-off strategies between detection accuracy and processing speed,the following strategies were deployed to develop the detector:feature extractor of inception V2,the model trained with common objects in context dataset,fixed_shape_resizer with the size of 600×600 pixels,kernel stride of 8300 proposals,and dynamic learning rate.The final detector had 95.7%recall,94.2%average precision,94.9%F1 score,23.5 mm root mean square error,and 8.3 fps processing speed,indicating decent performance for detecting poultry feeding behaviors.With the trained detector,temporal and spatial feeding behaviors of individual poultry can be successfully characterized.It is concluded that the faster RCNN should be a useful tool to continuously monitor poultry feeding behaviors in group settings.
