This article takes the joint training model as the research object and focuses on the joint postgraduate training in vocational universities.Based on a brief overview of the connotation and significance of joint postg...This article takes the joint training model as the research object and focuses on the joint postgraduate training in vocational universities.Based on a brief overview of the connotation and significance of joint postgraduate training in vocational universities,a systematic summary of the current situation of such training is conducted through practical research.While elaborating on the implementation and characteristics of joint postgraduate training in vocational universities,the article emphasizes a systematic collection and analysis of existing problems.Finally,guided by these problems and considering the characteristics and issues of joint postgraduate training in vocational universities,several improvement strategies are proposed.It is emphasized that vocational universities need to actively optimize their management models,establish restraint mechanisms,and implement dynamic adjustments to achieve scientific management of joint postgraduate training.This approach aims to promote academic innovation and improve training quality simultaneously.展开更多
Opinion question machine reading comprehension(MRC)requires a machine to answer questions by analyzing corresponding passages.Compared with traditional MRC tasks where the answer to every question is a segment of text...Opinion question machine reading comprehension(MRC)requires a machine to answer questions by analyzing corresponding passages.Compared with traditional MRC tasks where the answer to every question is a segment of text in corresponding passages,opinion question MRC is more challenging because the answer to an opinion question may not appear in corresponding passages but needs to be deduced from multiple sentences.In this study,a novel framework based on neural networks is proposed to address such problems,in which a new hybrid embedding training method combining text features is used.Furthermore,extra attention and output layers which generate auxiliary losses are introduced to jointly train the stacked recurrent neural networks.To deal with imbalance of the dataset,irrelevancy of question and passage is used for data augmentation.Experimental results show that the proposed method achieves state-of-the-art performance.We are the biweekly champion in the opinion question MRC task in Artificial Intelligence Challenger 2018(AIC2018).展开更多
Existing unsupervised person re-identification approaches fail to fully capture thefine-grained features of local regions,which can result in people with similar appearances and different identities being assigned the...Existing unsupervised person re-identification approaches fail to fully capture thefine-grained features of local regions,which can result in people with similar appearances and different identities being assigned the same label after clustering.The identity-independent information contained in different local regions leads to different levels of local noise.To address these challenges,joint training with local soft attention and dual cross-neighbor label smoothing(DCLS)is proposed in this study.First,the joint training is divided into global and local parts,whereby a soft attention mechanism is proposed for the local branch to accurately capture the subtle differences in local regions,which improves the ability of the re-identification model in identifying a person’s local significant features.Second,DCLS is designed to progressively mitigate label noise in different local regions.The DCLS uses global and local similarity metrics to semantically align the global and local regions of the person and further determines the proximity association between local regions through the cross information of neighboring regions,thereby achieving label smoothing of the global and local regions throughout the training process.In extensive experiments,the proposed method outperformed existing methods under unsupervised settings on several standard person re-identification datasets.展开更多
Intent detection and slot filling are two important components of natural language understanding.Because their relevance,joint training is often performed to improve performance.Existing studies mostly use a joint mod...Intent detection and slot filling are two important components of natural language understanding.Because their relevance,joint training is often performed to improve performance.Existing studies mostly use a joint model of multi-intent detection and slot-filling with unidirectional interaction,which improves the overall performance of the model by fusing the intent information in the slot-filling part.On this basis,in order to further improve the overall performance of the model by exploiting the correlation between the two,this paper proposes a joint multi-intent detection and slot-filling model based on a bidirectional interaction structure,which fuses the intent encoding information in the encoding part of slot filling and fuses the slot decoding information in the decoding part of intent detection.Experimental results on two public multi-intent joint training datasets,MixATIS and MixSNIPS,show that the bidirectional interaction structure proposed in this paper can effectively improve the performance of the joint model.In addition,in order to verify the generalization of the bidirectional interaction structure between intent and slot,a joint model for single-intent scenarios is proposed on the basis of the model in this paper.This model also achieves excellent performance on two public single-intent joint training datasets,CAIS and SNIPS.展开更多
Automatic speech recognition(ASR)systems have emerged as indispensable tools across a wide spectrum of applications,ranging from transcription services to voice-activated assistants.To enhance the performance of these...Automatic speech recognition(ASR)systems have emerged as indispensable tools across a wide spectrum of applications,ranging from transcription services to voice-activated assistants.To enhance the performance of these systems,it is important to deploy efficient models capable of adapting to diverse deployment conditions.In recent years,on-demand pruning methods have obtained significant attention within the ASR domain due to their adaptability in various deployment scenarios.However,these methods often confront substantial trade-offs,particularly in terms of unstable accuracy when reducing the model size.To address challenges,this study introduces two crucial empirical findings.Firstly,it proposes the incorporation of an online distillation mechanism during on-demand pruning training,which holds the promise of maintaining more consistent accuracy levels.Secondly,it proposes the utilization of the Mogrifier long short-term memory(LSTM)language model(LM),an advanced iteration of the conventional LSTM LM,as an effective alternative for pruning targets within the ASR framework.Through rigorous experimentation on the ASR system,employing the Mogrifier LSTM LM and training it using the suggested joint on-demand pruning and online distillation method,this study provides compelling evidence.The results exhibit that the proposed methods significantly outperform a benchmark model trained solely with on-demand pruning methods.Impressively,the proposed strategic configuration successfully reduces the parameter count by approximately 39%,all the while minimizing trade-offs.展开更多
In this paper, we introduce a novel scheme for the separate training of deep learning-based autoencoders used for Channel State Information (CSI) feedback. Our distinct training approach caters to multiple users and b...In this paper, we introduce a novel scheme for the separate training of deep learning-based autoencoders used for Channel State Information (CSI) feedback. Our distinct training approach caters to multiple users and base stations, enabling independent and individualized local training. This ensures the more secure processing of data and algorithms, different from the commonly adopted joint training method. To maintain comparable performance with joint training, we present two distinct training methods: separate training decoder and separate training encoder. It’s noteworthy that conducting separate training for the encoder can pose additional challenges, due to its responsibility in acquiring a compressed representation of underlying data features. This complexity makes accommodating multiple pre-trained decoders for just one encoder a demanding task. To overcome this, we design an adaptation layer architecture that effectively minimizes performance losses. Moreover, the flexible training strategy empowers users and base stations to seamlessly incorporate distinct encoder and decoder structures into the system, significantly amplifying the system’s scalability. .展开更多
Automation advancements prompts the extensive integration of collaborative robot(cobot)across a range of industries.Compared to the commonly used design approach of increasing the payload-to-weight ratio of cobot to e...Automation advancements prompts the extensive integration of collaborative robot(cobot)across a range of industries.Compared to the commonly used design approach of increasing the payload-to-weight ratio of cobot to enhance load capacity,equal attention should be paid to the dynamic response characteristics of cobot during the design process to make the cobot more flexible.In this paper,a new method for designing the drive train parameters of cobot is proposed.Firstly,based on the analysis of factors influencing the load capacity and dynamic response characteristics,design criteria for both aspects are established for cobot with all optimization design criteria normalized within the design domain.Secondly,with the cobot in the horizontal pose,the motor design scheme is discretized and it takes the joint motor diameter and gearbox speed ratio as optimization design variables.Finally,all the discrete values of the optimization objectives are obtained through the enumeration method and the Pareto front is used to select the optimal solution through multi-objective optimization.Base on the cobot design method proposed in this paper,a six-axis cobot is designed and compared with the commercial cobot.The result shows that the load capacity of the designed cobot in this paper reaches 8.4 kg,surpassing the 5 kg load capacity commercial cobot which is used as a benchmark.The minimum resonance frequency of the joints is 42.70 Hz.展开更多
Background:Marginal changes in the execution of competitive sports movements can represent a significant change for performance success.However,such differences may emerge only at certain execution intensities and are...Background:Marginal changes in the execution of competitive sports movements can represent a significant change for performance success.However,such differences may emerge only at certain execution intensities and are not easily detectable through conventional biomechanical techniques.This study aimed to investigate if and how competition standard and progression speed affect race walking kinematics from both a conventional and a coordination variability perspective.Methods:Fifteen experienced athletes divided into three groups(elite,international,and national) were studied while race walking on a treadmill at two different speeds(12.0 and 15.5 km/h).Basic gait parameters,the angular displacement of the pelvis and lower limbs,and the variability in continuous relative phase between six different joint couplings were analyzed.Results:Most of the spatio-temporal,kinematic,and coordination variability measures proved sensitive to the change in speed.Conversely,non-linear dynamics measures highlighted differences between athletes of different competition standard when conventional analytical tools were not able to discriminate between different skill levels.Continuous relative phase variability was higher for national level athletes than international and elite in two couplings(pelvis obliquity—hip flex/extension and pelvis rotation—ankle dorsi/plantarflexion) and gait phases(early stance for the first coupling,propulsive phase for the second) that are deemed fundamental for correct technique and performance.Conclusion:Measures of coordination variability showed to be a more sensitive tool for the fine detection of skill-dependent factors in competitive race walking,and showed good potential for being integrated in the assessment and monitoring of sports motor abilities.展开更多
The number of people with chronic diseases rises rapidly in recent years worldwide. Except for drug medication, mind-body exercises are indispensable for chronic disease management. Traditional Chinese practice （TCP...The number of people with chronic diseases rises rapidly in recent years worldwide. Except for drug medication, mind-body exercises are indispensable for chronic disease management. Traditional Chinese practice （TCP）, as an integrative intervention, is known as an effective means to keep in good health and fitness, as well as help regulate emotion. This paper introduces the domestic and overseas studies on effectiveness of TCP for chronic diseases, and explores the key action links from three aspects, including functional training of multiple-joint guided by consciousness, relieving psychological risk factors, improving respiratory and digestive function, blood and lymph circulation through respiratory training, and regulation of nerve, metabolic, and immune system. Finally, the authors discussed how to integrate TCP in the chronic disease management, and put forward that the practice methods and evaluation standard should be assessed academically.展开更多
Vehicle detection in dim light has always been a challenging task.In addition to the unavoidable noise,the uneven spatial distribution of light and dark due to vehicle lights and street lamps can further make the prob...Vehicle detection in dim light has always been a challenging task.In addition to the unavoidable noise,the uneven spatial distribution of light and dark due to vehicle lights and street lamps can further make the problem more difficult.Conventional image enhancement methods may produce over smoothing or over exposure problems,causing irreversible information loss to the vehicle targets to be subsequently detected.Therefore,we propose a multi-exposure generation and fusion network.In the multi-exposure generation network,we employ a single gated convolutional recurrent network with two-stream progressive exposure input to generate intermediate images with gradually increasing exposure,which are provided to the multi-exposure fusion network after a spatial attention mechanism.Then,a pre-trained vehicle detection model in normal light is used as the basis of the fusion network,and the two models are connected using the convolutional kernel channel dimension expansion technique.This allows the fusion module to provide vehicle detection information,which can be used to guide the generation network tofine-tune the parameters and thus complete end-to-end enhancement and training.By coupling the two parts,we can achieve detail interaction and feature fusion under different lighting conditions.Our experimental results demonstrate that our proposed method is better than the state-of-the-art detection methods after image luminance enhancement on the ODDS dataset.展开更多
基金SZPU Quality Engineering Project University-level Teaching and Research Project(2025-29),Exploration and Practice of Professional Degree Cultivation and Joint Graduate Training in Vocational Undergraduate CollegesThe Young Backbone Teacher Plan of Beijing Information Science and Technology University(Project No.:YBT202448)+1 种基金Industry-University Cooperation Education Project of the Ministry of Education,Research on the Application of Model-Based Viewpoint Selection in UAV 3D Modeling Course Teaching(Project No.:1007-7025210681J1)Research on the Path of Promoting the Reform of Scientific Research Evaluation in Vocational Undergraduate University(Project No.:GDKG2025ZC1)。
文摘This article takes the joint training model as the research object and focuses on the joint postgraduate training in vocational universities.Based on a brief overview of the connotation and significance of joint postgraduate training in vocational universities,a systematic summary of the current situation of such training is conducted through practical research.While elaborating on the implementation and characteristics of joint postgraduate training in vocational universities,the article emphasizes a systematic collection and analysis of existing problems.Finally,guided by these problems and considering the characteristics and issues of joint postgraduate training in vocational universities,several improvement strategies are proposed.It is emphasized that vocational universities need to actively optimize their management models,establish restraint mechanisms,and implement dynamic adjustments to achieve scientific management of joint postgraduate training.This approach aims to promote academic innovation and improve training quality simultaneously.
基金Project supported by the China Knowledge Centre for Engineering Sciences and Technology(No.CKCEST-2019-1-12)the National Natural Science Foundation of China(No.61572434)。
文摘Opinion question machine reading comprehension(MRC)requires a machine to answer questions by analyzing corresponding passages.Compared with traditional MRC tasks where the answer to every question is a segment of text in corresponding passages,opinion question MRC is more challenging because the answer to an opinion question may not appear in corresponding passages but needs to be deduced from multiple sentences.In this study,a novel framework based on neural networks is proposed to address such problems,in which a new hybrid embedding training method combining text features is used.Furthermore,extra attention and output layers which generate auxiliary losses are introduced to jointly train the stacked recurrent neural networks.To deal with imbalance of the dataset,irrelevancy of question and passage is used for data augmentation.Experimental results show that the proposed method achieves state-of-the-art performance.We are the biweekly champion in the opinion question MRC task in Artificial Intelligence Challenger 2018(AIC2018).
基金supported by the National Natural Science Foundation of China under Grant Nos.62076117 and 62166026the Jiangxi Key Laboratory of Smart City under Grant No.20192BCD40002Jiangxi Provincial Natural Science Foundation under Grant No.20224BAB212011.
文摘Existing unsupervised person re-identification approaches fail to fully capture thefine-grained features of local regions,which can result in people with similar appearances and different identities being assigned the same label after clustering.The identity-independent information contained in different local regions leads to different levels of local noise.To address these challenges,joint training with local soft attention and dual cross-neighbor label smoothing(DCLS)is proposed in this study.First,the joint training is divided into global and local parts,whereby a soft attention mechanism is proposed for the local branch to accurately capture the subtle differences in local regions,which improves the ability of the re-identification model in identifying a person’s local significant features.Second,DCLS is designed to progressively mitigate label noise in different local regions.The DCLS uses global and local similarity metrics to semantically align the global and local regions of the person and further determines the proximity association between local regions through the cross information of neighboring regions,thereby achieving label smoothing of the global and local regions throughout the training process.In extensive experiments,the proposed method outperformed existing methods under unsupervised settings on several standard person re-identification datasets.
基金Supported by the National Nature Science Foundation of China(62462037,62462036)Project for Academic and Technical Leader in Major Disciplines in Jiangxi Province(20232BCJ22013)+1 种基金Jiangxi Provincial Natural Science Foundation(20242BAB26017,20232BAB202010)Jiangxi Province Graduate Innovation Fund Project(YC2023-S320)。
文摘Intent detection and slot filling are two important components of natural language understanding.Because their relevance,joint training is often performed to improve performance.Existing studies mostly use a joint model of multi-intent detection and slot-filling with unidirectional interaction,which improves the overall performance of the model by fusing the intent information in the slot-filling part.On this basis,in order to further improve the overall performance of the model by exploiting the correlation between the two,this paper proposes a joint multi-intent detection and slot-filling model based on a bidirectional interaction structure,which fuses the intent encoding information in the encoding part of slot filling and fuses the slot decoding information in the decoding part of intent detection.Experimental results on two public multi-intent joint training datasets,MixATIS and MixSNIPS,show that the bidirectional interaction structure proposed in this paper can effectively improve the performance of the joint model.In addition,in order to verify the generalization of the bidirectional interaction structure between intent and slot,a joint model for single-intent scenarios is proposed on the basis of the model in this paper.This model also achieves excellent performance on two public single-intent joint training datasets,CAIS and SNIPS.
基金supported by Institute of Information&communications Technology Planning&Evaluation(IITP)grant funded by the Korea government(MSIT)(No.2022-0-00377,Development of Intelligent Analysis and Classification Based Contents Class Categorization Technique to Prevent Imprudent Harmful Media Distribution).
文摘Automatic speech recognition(ASR)systems have emerged as indispensable tools across a wide spectrum of applications,ranging from transcription services to voice-activated assistants.To enhance the performance of these systems,it is important to deploy efficient models capable of adapting to diverse deployment conditions.In recent years,on-demand pruning methods have obtained significant attention within the ASR domain due to their adaptability in various deployment scenarios.However,these methods often confront substantial trade-offs,particularly in terms of unstable accuracy when reducing the model size.To address challenges,this study introduces two crucial empirical findings.Firstly,it proposes the incorporation of an online distillation mechanism during on-demand pruning training,which holds the promise of maintaining more consistent accuracy levels.Secondly,it proposes the utilization of the Mogrifier long short-term memory(LSTM)language model(LM),an advanced iteration of the conventional LSTM LM,as an effective alternative for pruning targets within the ASR framework.Through rigorous experimentation on the ASR system,employing the Mogrifier LSTM LM and training it using the suggested joint on-demand pruning and online distillation method,this study provides compelling evidence.The results exhibit that the proposed methods significantly outperform a benchmark model trained solely with on-demand pruning methods.Impressively,the proposed strategic configuration successfully reduces the parameter count by approximately 39%,all the while minimizing trade-offs.
文摘In this paper, we introduce a novel scheme for the separate training of deep learning-based autoencoders used for Channel State Information (CSI) feedback. Our distinct training approach caters to multiple users and base stations, enabling independent and individualized local training. This ensures the more secure processing of data and algorithms, different from the commonly adopted joint training method. To maintain comparable performance with joint training, we present two distinct training methods: separate training decoder and separate training encoder. It’s noteworthy that conducting separate training for the encoder can pose additional challenges, due to its responsibility in acquiring a compressed representation of underlying data features. This complexity makes accommodating multiple pre-trained decoders for just one encoder a demanding task. To overcome this, we design an adaptation layer architecture that effectively minimizes performance losses. Moreover, the flexible training strategy empowers users and base stations to seamlessly incorporate distinct encoder and decoder structures into the system, significantly amplifying the system’s scalability. .
基金Supported by National Key Research and Development Program of China (Grant Nos.2022YFB4703000,2019YFB1309900)。
文摘Automation advancements prompts the extensive integration of collaborative robot(cobot)across a range of industries.Compared to the commonly used design approach of increasing the payload-to-weight ratio of cobot to enhance load capacity,equal attention should be paid to the dynamic response characteristics of cobot during the design process to make the cobot more flexible.In this paper,a new method for designing the drive train parameters of cobot is proposed.Firstly,based on the analysis of factors influencing the load capacity and dynamic response characteristics,design criteria for both aspects are established for cobot with all optimization design criteria normalized within the design domain.Secondly,with the cobot in the horizontal pose,the motor design scheme is discretized and it takes the joint motor diameter and gearbox speed ratio as optimization design variables.Finally,all the discrete values of the optimization objectives are obtained through the enumeration method and the Pareto front is used to select the optimal solution through multi-objective optimization.Base on the cobot design method proposed in this paper,a six-axis cobot is designed and compared with the commercial cobot.The result shows that the load capacity of the designed cobot in this paper reaches 8.4 kg,surpassing the 5 kg load capacity commercial cobot which is used as a benchmark.The minimum resonance frequency of the joints is 42.70 Hz.
文摘Background:Marginal changes in the execution of competitive sports movements can represent a significant change for performance success.However,such differences may emerge only at certain execution intensities and are not easily detectable through conventional biomechanical techniques.This study aimed to investigate if and how competition standard and progression speed affect race walking kinematics from both a conventional and a coordination variability perspective.Methods:Fifteen experienced athletes divided into three groups(elite,international,and national) were studied while race walking on a treadmill at two different speeds(12.0 and 15.5 km/h).Basic gait parameters,the angular displacement of the pelvis and lower limbs,and the variability in continuous relative phase between six different joint couplings were analyzed.Results:Most of the spatio-temporal,kinematic,and coordination variability measures proved sensitive to the change in speed.Conversely,non-linear dynamics measures highlighted differences between athletes of different competition standard when conventional analytical tools were not able to discriminate between different skill levels.Continuous relative phase variability was higher for national level athletes than international and elite in two couplings(pelvis obliquity—hip flex/extension and pelvis rotation—ankle dorsi/plantarflexion) and gait phases(early stance for the first coupling,propulsive phase for the second) that are deemed fundamental for correct technique and performance.Conclusion:Measures of coordination variability showed to be a more sensitive tool for the fine detection of skill-dependent factors in competitive race walking,and showed good potential for being integrated in the assessment and monitoring of sports motor abilities.
文摘The number of people with chronic diseases rises rapidly in recent years worldwide. Except for drug medication, mind-body exercises are indispensable for chronic disease management. Traditional Chinese practice （TCP）, as an integrative intervention, is known as an effective means to keep in good health and fitness, as well as help regulate emotion. This paper introduces the domestic and overseas studies on effectiveness of TCP for chronic diseases, and explores the key action links from three aspects, including functional training of multiple-joint guided by consciousness, relieving psychological risk factors, improving respiratory and digestive function, blood and lymph circulation through respiratory training, and regulation of nerve, metabolic, and immune system. Finally, the authors discussed how to integrate TCP in the chronic disease management, and put forward that the practice methods and evaluation standard should be assessed academically.
基金supported in part by the Science and Technology Innovation foundation(No.JSGG20210802152811033).
文摘Vehicle detection in dim light has always been a challenging task.In addition to the unavoidable noise,the uneven spatial distribution of light and dark due to vehicle lights and street lamps can further make the problem more difficult.Conventional image enhancement methods may produce over smoothing or over exposure problems,causing irreversible information loss to the vehicle targets to be subsequently detected.Therefore,we propose a multi-exposure generation and fusion network.In the multi-exposure generation network,we employ a single gated convolutional recurrent network with two-stream progressive exposure input to generate intermediate images with gradually increasing exposure,which are provided to the multi-exposure fusion network after a spatial attention mechanism.Then,a pre-trained vehicle detection model in normal light is used as the basis of the fusion network,and the two models are connected using the convolutional kernel channel dimension expansion technique.This allows the fusion module to provide vehicle detection information,which can be used to guide the generation network tofine-tune the parameters and thus complete end-to-end enhancement and training.By coupling the two parts,we can achieve detail interaction and feature fusion under different lighting conditions.Our experimental results demonstrate that our proposed method is better than the state-of-the-art detection methods after image luminance enhancement on the ODDS dataset.
