摘要 :
In a growing number of information processing applications, data takes the form of continuous data streams rather than traditional stored databases. Most of these applications have sophisticated real-time constraint that needs to ...
展开
In a growing number of information processing applications, data takes the form of continuous data streams rather than traditional stored databases. Most of these applications have sophisticated real-time constraint that needs to be met under unbounded, high-volume, time-varying data streams. We introduce deadline as a real-time constraint of continuous query over data streams. Specifically, a deadline-sensitive approach for sliding window processing is proposed, which predicts a sliding window would satisfy its deadline or not. Once deadline missing is predicted, HopDrop, a proposed load shedding strategy would be applied to lighten the system burden in advance. Furthermore, feedback control mechanism is applied to improve the adaptivity of our approach. Extensive experimental results are presented and analyzed to validate our strategies.
收起
摘要 :
In a growing number of information processing applications, data takes the form of continuous data streams rather than traditional stored databases. Most of these applications have sophisticated real-time constraint that needs to ...
展开
In a growing number of information processing applications, data takes the form of continuous data streams rather than traditional stored databases. Most of these applications have sophisticated real-time constraint that needs to be met under unbounded, high-volume, time-varying data streams. We introduce deadline as a real-time constraint of continuous query over data streams. Specifically, a deadline-sensitive approach for sliding window processing is proposed, which predicts a sliding window would satisfy its deadline or not. Once deadline missing is predicted, HopDrop, a proposed load shedding strategy would be applied to lighten the system burden in advance. Furthermore, feedback control mechanism is applied to improve the adaptivity of our approach. Extensive experimental results are presented and analyzed to validate our strategies.
收起
摘要 :
This paper presents THU-SOC, a new methodology and tool dedicated to explore design space by executing full-scale SW application code on the transaction level models of the SoC platform. The SoC platform supports alternative Trans...
展开
This paper presents THU-SOC, a new methodology and tool dedicated to explore design space by executing full-scale SW application code on the transaction level models of the SoC platform. The SoC platform supports alternative Transaction Level Models (TLMs), bus-functional model and bus-arbitration model, which enables it to cooperate with different levels of hardware descriptions. So, users are only required to provide functional descriptions to construct a whole cycle-accurate system simulation for a broad design space exploration in the architecture design. When the architecture is determined, the high-level descriptions can be replaced by RTL-level descriptions to accomplish the system verification, and the interface between the tool and the descriptions is unmodified. Moreover, THU-SOC integrates some behavior models of necessary components in a SoC system, such as ISS (Instruction-Set Simulator) simulator of CPU, interrupt controller, bus arbiter, memory controller, UART controller, so users can focus themselves on the design of the target component. The tool is written in C++ and supports the PLI (Programming Language Interface), therefore its performance is satisfying and different kinds of hardware description languages, such as System-C, Verilog, VHDL and so on, are supported.
收起
摘要 :
This paper presents THU-SOC, a new methodology and tool dedicated to explore design space by executing full-scale SW application code on the transaction level models of the SoC platform. The SoC platform supports alternative Trans...
展开
This paper presents THU-SOC, a new methodology and tool dedicated to explore design space by executing full-scale SW application code on the transaction level models of the SoC platform. The SoC platform supports alternative Transaction Level Models (TLMs), bus-functional model and bus-arbitration model, which enables it to cooperate with different levels of hardware descriptions. So, users are only required to provide functional descriptions to construct a whole cycle-accurate system simulation for a broad design space exploration in the architecture design. When the architecture is determined, the high-level descriptions can be replaced by RTL-level descriptions to accomplish the system verification, and the interface between the tool and the descriptions is unmodified. Moreover, THU-SOC integrates some behavior models of necessary components in a SoC system, such as ISS (Instruction-Set Simulator) simulator of CPU, interrupt controller, bus arbiter, memory controller, UART controller, so users can focus themselves on the design of the target component. The tool is written in C++ and supports the PLI (Programming Language Interface), therefore its performance is satisfying and different kinds of hardware description languages, such as System-C, Verilog, VHDL and so on, are supported.
收起
摘要 :
Human pulmonary function declines with age. Elders, especially those with lung or cardiovascular diseases, yearn for daily lung function tests for timely diagnosis and treatment. However, current clinical spirometers are cumbersom...
展开
Human pulmonary function declines with age. Elders, especially those with lung or cardiovascular diseases, yearn for daily lung function tests for timely diagnosis and treatment. However, current clinical spirometers are cumbersome and ex-pensive while home-use portable ones’ accuracy is questionable. Moreover, both kinds require contact measurements and could cause cross infection, especially hazardous for contagious diseases like COVID-19. To this end, we propose SpiroFi, a contactless system that leverages WiFi Channel State Information (CSI) for convenient yet accurate Pulmonary Function Testing (PFT) out of clinic. The key enabler underlying SpiroFi is a set of algorithms that can extract chest wall movement from WiFi signal variations and interpret such information into lung function indices. We have realized SpiroFi on low-cost commodity WiFi devices and tested it in a home-like site where it achieves 2.55% monitoring error over healthy youths. Then, with the Ethics Committee (EC) approval, we conducted a 2-month clinic study in a city hospital over elders with basic diseases. SprioFi still yields 6.05% monitoring error despite elders’ degenerated pulmonary function and body control. Also, the correlation between lung function and age as well as chronic diseases has been revealed, highlighting the importance of daily PFT for the elderly.
收起
摘要 :
The aim of the paper is to examine the relationship between corporate governance and R&D expenditure of firm-level in China. It also analyzes the mechanisms of corporate governance's impact on R&D investments. On the basis of the former research in the field of corporate governance and R&D investments, the paper explains three aspects of corporate governance and investment financing of R&D: the nature of the principle, the extent of ownership and control, and characteristic of the board. The paper proposes three hypotheses and analyzes the possible pathway corporate governance works on R&D investment. In order to test the hypotheses above, we set up the statistical to do the empirical analysis and get related results. And we conclude some useful implications for future corporation's performance and policy....
展开
The aim of the paper is to examine the relationship between corporate governance and R&D expenditure of firm-level in China. It also analyzes the mechanisms of corporate governance's impact on R&D investments. On the basis of the former research in the field of corporate governance and R&D investments, the paper explains three aspects of corporate governance and investment financing of R&D: the nature of the principle, the extent of ownership and control, and characteristic of the board. The paper proposes three hypotheses and analyzes the possible pathway corporate governance works on R&D investment. In order to test the hypotheses above, we set up the statistical to do the empirical analysis and get related results. And we conclude some useful implications for future corporation's performance and policy.
收起
摘要 :
Due to limited training data for SAR target recognition tasks, different regularization techniques have been used in the deep convolution neural networks (CNN) to improve generalization ability. In this paper, the influences of th...
展开
Due to limited training data for SAR target recognition tasks, different regularization techniques have been used in the deep convolution neural networks (CNN) to improve generalization ability. In this paper, the influences of three regularization techniques, including data augmentation, L_2 regularization term, dropout, are studied under standard operating conditions (SOC) when moving and stationary target recognition (MSTAR) dataset is used for SAR target recognition. Four representative CNN models based on classical models, such as AlexNet and ResNet, are selected and trained to recognize 10-classes targets. Additionally, a CNN model which has fewer network parameters is designed based on multi-scale spatial feature extraction strategy and SqueezeNet to study the influence of the amount of network parameters. The experimental results demonstrate that, when using the AlexNet series model for SAR target recognition, using dropout may greatly improve the ability of model optimization. ResNet series models which have more layers, have better effect on Test 1+noise than other CNN models, especially taking dropout in the model. For the models based on highway networks, adding L_2 regularization terms in loss function can improve the test accuracy, but it also makes the latter phase of training extremely unstable. Data augmentation is an effective regularization technique when the model can get high training accuracy.
收起
摘要 :
Due to limited training data for SAR target recognition tasks, different regularization techniques have been used in
the deep convolution neural networks (CNN) to improve generalization ability. In this paper, the influences of th...
展开
Due to limited training data for SAR target recognition tasks, different regularization techniques have been used in
the deep convolution neural networks (CNN) to improve generalization ability. In this paper, the influences of three
regularization techniques, including data augmentation, L_2 regularization term, dropout, are studied under standard
operating conditions (SOC) when moving and stationary target recognition (MSTAR) dataset is used for SAR target
recognition. Four representative CNN models based on classical models, such as AlexNet and ResNet, are selected and
trained to recognize 10-classes targets. Additionally, a CNN model which has fewer network parameters is designed
based on multi-scale spatial feature extraction strategy and SqueezeNet to study the influence of the amount of network
parameters. The experimental results demonstrate that, when using the AlexNet series model for SAR target recognition,
using dropout may greatly improve the ability of model optimization. ResNet series models which have more layers,
have better effect on Test 1+noise than other CNN models, especially taking dropout in the model. For the models based
on highway networks, adding L_2 regularization terms in loss function can improve the test accuracy, but it also makes the
latter phase of training extremely unstable. Data augmentation is an effective regularization technique when the model
can get high training accuracy.
收起
摘要 :
Due to limited training data for SAR target recognition tasks, different regularization techniques have been used in
the deep convolution neural networks (CNN) to improve generalization ability. In this paper, the influences of t...
展开
Due to limited training data for SAR target recognition tasks, different regularization techniques have been used in
the deep convolution neural networks (CNN) to improve generalization ability. In this paper, the influences of three
regularization techniques, including data augmentation, L_2 regularization term, dropout, are studied under standard
operating conditions (SOC) when moving and stationary target recognition (MSTAR) dataset is used for SAR target
recognition. Four representative CNN models based on classical models, such as AlexNet and ResNet, are selected and
trained to recognize 10-classes targets. Additionally, a CNN model which has fewer network parameters is designed
based on multi-scale spatial feature extraction strategy and SqueezeNet to study the influence of the amount of network
parameters. The experimental results demonstrate that, when using the AlexNet series model for SAR target recognition,
using dropout may greatly improve the ability of model optimization. ResNet series models which have more layers,
have better effect on Test 1+noise than other CNN models, especially taking dropout in the model. For the models based
on highway networks, adding L_2 regularization terms in loss function can improve the test accuracy, but it also makes the
latter phase of training extremely unstable. Data augmentation is an effective regularization technique when the model
can get high training accuracy.
收起
摘要 :
Because of low precision of license plate location in the color vehicle images with complex background, a new License Plate Location Algorithm was proposed based on multi-scale edge detection method with wavelet transform. Firstly...
展开
Because of low precision of license plate location in the color vehicle images with complex background, a new License Plate Location Algorithm was proposed based on multi-scale edge detection method with wavelet transform. Firstly, we perform WT on two directions respectively, and obtain the magnitude of the 2-D WT. Then detect the maxima magnitude of WT as the image edge and get edge image. At last, the projection approach based on two-way back is adopted to examine up-down and left-right boundary of the car license. Experimental results demonstrate that the proposed method can enhance the images location speed and accurate capture license plate region.
收起