The deep convolutional neural network(CNN)is exploited in this work to conduct the challenging channel estimation for mmWave massive multiple input multiple output(MIMO)systems.The inherent sparse features of the mmWa...The deep convolutional neural network(CNN)is exploited in this work to conduct the challenging channel estimation for mmWave massive multiple input multiple output(MIMO)systems.The inherent sparse features of the mmWave massive MIMO channels can be extracted and the sparse channel supports can be learnt by the multi-layer CNN-based network through training.Then accurate channel inference can be efficiently implemented using the trained network.The estimation accuracy and spectrum efficiency can be further improved by fully utilizing the spatial correlation among the sparse channel supports of different antennas.It is verified by simulation results that the proposed deep CNN-based scheme significantly outperforms the state-of-the-art benchmarks in both accuracy and spectrum efficiency.展开更多
In urban Vehicular Ad hoc Networks(VANETs),high mobility of vehicular environment and frequently changed network topology call for a low delay end-to-end routing algorithm.In this paper,we propose a Multi-Agent Reinfo...In urban Vehicular Ad hoc Networks(VANETs),high mobility of vehicular environment and frequently changed network topology call for a low delay end-to-end routing algorithm.In this paper,we propose a Multi-Agent Reinforcement Learning(MARL)based decentralized routing scheme,where the inherent similarity between the routing problem in VANET and the MARL problem is exploited.The proposed routing scheme models the interaction between vehicles and the environment as a multi-agent problem in which each vehicle autonomously establishes the communication channel with a neighbor device regardless of the global information.Simulation performed in the 3GPP Manhattan mobility model demonstrates that our proposed decentralized routing algorithm achieves less than 45.8 ms average latency and high stability of 0.05%averaging failure rate with varying vehicle capacities.展开更多
It has always been difficult to achieve accurate information of the channel for underwater acoustic communications because of the severe underwater propagation conditions,including frequency-selective property,high re...It has always been difficult to achieve accurate information of the channel for underwater acoustic communications because of the severe underwater propagation conditions,including frequency-selective property,high relative mobility,long propagation latency,and intensive ambient noise,etc.To this end,a deep unfolding neural network based approach is proposed,in which multiple layers of the network mimic the iterations of the classical iterative sparse approximation algorithm to extract the inherent sparse features of the channel by exploiting deep learning,and a scheme based on the Sparsity-Aware DNN(SA-DNN)for UAC estimation is proposed to improve the estimation accuracy.Moreover,we propose a Denoising Sparsity-Aware DNN(DeSA-DNN)based enhanced method that integrates a denoising CNN module in the sparsity-aware deep network,so that the degradation brought by intensive ambient noise could be eliminated and the estimation accuracy can be further improved.Simulation results demonstrate that the performance of the proposed schemes is superior to the state-of-the-art compressed sensing based and iterative sparse recovery schems in the aspects of channel recovery precision,pilot overhead,and robustness,particularly under unideal circumstances of intensive ambient noise or inadequate measurement pilots.展开更多
Despite the rapid development of mobile and embedded hardware, directly executing computationexpensive and storage-intensive deep learning algorithms on these devices’ local side remains constrained for sensory data ...Despite the rapid development of mobile and embedded hardware, directly executing computationexpensive and storage-intensive deep learning algorithms on these devices’ local side remains constrained for sensory data analysis. In this paper, we first summarize the layer compression techniques for the state-of-theart deep learning model from three categories: weight factorization and pruning, convolution decomposition, and special layer architecture designing. For each category of layer compression techniques, we quantify their storage and computation tunable by layer compression techniques and discuss their practical challenges and possible improvements. Then, we implement Android projects using TensorFlow Mobile to test these 10 compression methods and compare their practical performances in terms of accuracy, parameter size, intermediate feature size,computation, processing latency, and energy consumption. To further discuss their advantages and bottlenecks,we test their performance over four standard recognition tasks on six resource-constrained Android smartphones.Finally, we survey two types of run-time Neural Network(NN) compression techniques which are orthogonal with the layer compression techniques, run-time resource management and cost optimization with special NN architecture,which are orthogonal with the layer compression techniques.展开更多
Lack of physical activity is becoming a killer of our healthy life. As a solution for this negative impact,we propose Smart Care to help users to set up a healthy physical activity habit. Smart Care can monitor a user...Lack of physical activity is becoming a killer of our healthy life. As a solution for this negative impact,we propose Smart Care to help users to set up a healthy physical activity habit. Smart Care can monitor a user's activities over a long time, and then provide activity quality assessment and suggestion. Smart Care consists of three parts, activity recognition, energy saving, and health feedback. Activity recognition can recognize nine kinds of daily activities. A hybrid classifier that uses less power and memory with satisfactory accuracy was designed and implemented by utilizing the periodicity of target activity. In addition, a learning-based energy saver was introduced to reduce energy consumption by adjusting sampling rates and the set of features adaptively. Based on the type and duration of the activity recorded, health feedback in terms of the calorie burned was given. The system could provide quantitative activity quality assessment and recommend future physical activity plans. Through extensive real-life testing, the system is shown to achieve an average recognition accuracy of 98.0% with a minimized energy expenditure.展开更多
A single output Q-switched Nd:GdVO4 laser with a reflective graphene oxide(GO) saturable absorber was demonstrated. The shortest pulse duration in the Q-switched laser is 115 ns, and the output power ranges from1.23 W...A single output Q-switched Nd:GdVO4 laser with a reflective graphene oxide(GO) saturable absorber was demonstrated. The shortest pulse duration in the Q-switched laser is 115 ns, and the output power ranges from1.23 W at 1.71 MHz to 2.11 W at 2.50 MHz when the pump power rises from 7.40 to 10.90 W with the utilization of GO Langmuir–Blodgett(LB) films based on the convenient and low-cost LB technique. To the best of our knowledge, it is the highest output power in a Q-switched laser with a GO saturable absorber.展开更多
Conventional change detection approaches are mainly based on per-pixel processing,which ignore the sub-pixel spectral variation resulted from spectral mixture.Especially for medium-resolution remote sensing images use...Conventional change detection approaches are mainly based on per-pixel processing,which ignore the sub-pixel spectral variation resulted from spectral mixture.Especially for medium-resolution remote sensing images used in urban landcover change monitoring,land use/cover components within a single pixel are usually complicated and heterogeneous due to the limitation of the spatial resolution.Thus,traditional hard detection methods based on pure pixel assumption may lead to a high level of omission and commission errors inevitably,degrading the overall accuracy of change detection.In order to address this issue and find a possible way to exploit the spectral variation in a sub-pixel level,a novel change detection scheme is designed based on the spectral mixture analysis and decision-level fusion.Nonlinear spectral mixture model is selected for spectral unmixing,and change detection is implemented in a sub-pixel level by investigating the inner-pixel subtle changes and combining multiple composition evidences.The proposed method is tested on multi-temporal Landsat Thematic Mapper and China–Brazil Earth Resources Satellite remote sensing images for the land-cover change detection over urban areas.The effectiveness of the proposed approach is confirmed in terms of several accuracy indices in contrast with two pixel-based change detection methods(i.e.change vector analysis and principal component analysis-based method).In particular,the proposed sub-pixel change detection approach not only provides the binary change information,but also obtains the characterization about change direction and intensity,which greatly extends the semantic meaning of the detected change targets.展开更多
In this article, we report on an experimentally generated soliton and bound-state soliton passively mode-locked erbium-doped fiber laser by incorporating a saturable absorber(SA) made of MoS2∕fluorine mica(FM) that w...In this article, we report on an experimentally generated soliton and bound-state soliton passively mode-locked erbium-doped fiber laser by incorporating a saturable absorber(SA) made of MoS2∕fluorine mica(FM) that was fabricated with the Langmuir–Blodgett(LB) method. The FM substrate is 20 μm thick and easy to bend or cut,like a polymer. However, it has a higher damage threshold and a better thermal dissipation than polymers.In addition, the LB method can be used to fabricate a thin film with good uniformity. In this study, the modulation depth, saturable intensity, and unsaturated loss of the SA are measured as 5.9%, 57.69 MW∕cm2, and13.4%, respectively. Based on the SA, a soliton mode-locked laser is achieved. The pulse duration, repetition rate, and signal-to-noise ratio are 581 fs, 15.67 MHz, and 65 dB, respectively. By adjusting the polarization controller and pump power, we obtain a bound-state soliton mode-locked pulse. The temporal interval between the two solitons forming the bound-state pulse is 2.7 ps. The repetition rate of the bound-state pulses is proportional to the pump power. The maximum repetition rate is 517 MHz, corresponding to the 33 rd harmonic of the fundamental repetition rate. The results indicate that the MoS2∕FM LB film absorber is a promising photonic device in ultrafast fiber lasers.展开更多
The dissolution of collagen in ionic liquids(ILs)was highly dependent on the polarity of ILs,which was influenced by their sorts and concentrations.Herein,the solubility and dispersion degree of collagen in two sorts ...The dissolution of collagen in ionic liquids(ILs)was highly dependent on the polarity of ILs,which was influenced by their sorts and concentrations.Herein,the solubility and dispersion degree of collagen in two sorts of ILs,namely 1-ethyl-methylimidazolium tetrafluoroborate([EMIM][BF4])with low polarity and 1-ethyl-3-methylimidazolium acetate([EMIM][Ac])with high polarity in a concentration range from 10% to 70% at 10℃ were investigated.When 150 mg of collagen was added to 30 mg of ILs,the minimum soluble collagen concentration was 0.02 mg/mL in 70%[EMIM][BF4]with lowest polarity and the maximum was 3.57 mg/mL in 70%[EMIM][Ac]with highest polarity,which indicates that soluble collagen and insoluble collagen fibers were both present.For insoluble collagens,differential scanning calorimetry showed that the thermal-stability was weakened when increasing the ILs concentration and polarity,and the fiber arrangement was looser with a more uniform lyophilized structure,observed by atomic force microscopy and scanning electron microscopy.For soluble collagens,electrophoresis patterns and Fourier transform infrared spectroscopy showed that no polypeptide chain degradation occurred during dissolution,but the thermal denaturation temperature decreased by 0.26℃~7.63℃ with the increase of ILs concentrations,measured by ultra-sensitive differential scanning calorimetry.Moreover,the aggregation of collagen molecules was reduced when ILs polarity was increased as determined by fluorescence measurements and dynamic light scattering,which resulted in an increased loose fiber arrangement observed by atomic force microscopy.If the structural integrity of collagen needs to be retained,then the ILs sorts and concentrations should be considered.展开更多
1 Introduction Withthe rapid development of the E-commerce,more consumers turn to shop online.When online shoppers search for products using keywords,the related products appear with their brief descriptions.We conduc...1 Introduction Withthe rapid development of the E-commerce,more consumers turn to shop online.When online shoppers search for products using keywords,the related products appear with their brief descriptions.We conduct a real-world survey and find that most shoppers are dissatisfied with the existing"onefit-all product descriptions"and they have to spend more time to scan detail pages.However,handcrafting the attractive product descriptions is always costly.展开更多
Estates,especially those of public securityrelated companies and institutes,have to protect their privacy from adversary unmanned aerial vehicles(UAVs).In this paper,we propose a reinforcement learning-based control f...Estates,especially those of public securityrelated companies and institutes,have to protect their privacy from adversary unmanned aerial vehicles(UAVs).In this paper,we propose a reinforcement learning-based control framework to prevent unauthorized UAVs from entering a target area in a dynamic game without being aware of the UAV attack model.This UAV control scheme enables a target estate to choose the optimal control policy,such as jamming the global positioning system signals,hacking,and laser shooting,to expel nearby UAVs.A deep reinforcement learning technique,called neural episodic control,is used to accelerate the learning speed to achieve the optimal UAV control policy,especially for estates with a large area,against complicated UAV attack policies.We analyze the computational complexity for the proposed UAV control scheme and provide its performance bound,including the risk level of the estate and its utility.Our simulation results show that the proposed scheme can reduce the risk level of the target estate and improve its utility against malicious UAVs compared with the selected benchmark scheme.展开更多
The video transmission in the Internet-of-Things(IoT)system must guarantee the video quality and reduce the packet loss rate and the delay with limited resources to satisfy the requirement of multimedia services.In th...The video transmission in the Internet-of-Things(IoT)system must guarantee the video quality and reduce the packet loss rate and the delay with limited resources to satisfy the requirement of multimedia services.In this paper,we propose a reinforcement learning based energy-efficient IoT video transmission scheme that protects against interference,in which the base station controls the transmission action of the IoT device including the encoding rate,the modulation and coding scheme,and the transmit power.A reinforcement learning algorithm state-action-reward-state-action is applied to choose the transmission action based on the observed state(the queue length of the buffer,the channel gain,the previous bit error rate,and the previous packet loss rate)without knowledge of the transmission channel model at the transmitter and the receiver.We also propose a deep reinforcement learning based energy-efficient IoT video transmission scheme that uses a deep neural network to approximate Q value to further accelerate the learning process involved in choosing the optimal transmission action and improve the video transmission performance.Moreover,both the performance bounds of the proposed schemes and the computational complexity are theoretically derived.Simulation results show that the proposed schemes can increase the peak signal-to-noise ratio and decrease the packet loss rate,the delay,and the energy consumption relative to the benchmark scheme.展开更多
基金This work is supported in part by the National Natural Science Foundation of China under grants 61901403,61971366 and 61971365in part by the Youth Innovation Fund of Xiamen under grant 3502Z20206039in part by the Natural Science Foundation of Fujian Province of China under grant 2019J05001.
文摘The deep convolutional neural network(CNN)is exploited in this work to conduct the challenging channel estimation for mmWave massive multiple input multiple output(MIMO)systems.The inherent sparse features of the mmWave massive MIMO channels can be extracted and the sparse channel supports can be learnt by the multi-layer CNN-based network through training.Then accurate channel inference can be efficiently implemented using the trained network.The estimation accuracy and spectrum efficiency can be further improved by fully utilizing the spatial correlation among the sparse channel supports of different antennas.It is verified by simulation results that the proposed deep CNN-based scheme significantly outperforms the state-of-the-art benchmarks in both accuracy and spectrum efficiency.
基金This work is supported by the National Science Foundation of China under grant No.61901403,61790551,and 61925106,Youth Innovation Fund of Xiamen No.3502Z20206039 and Tsinghua-Foshan Innovation Special Fund(TFISF)No.2020THFS0109.
文摘In urban Vehicular Ad hoc Networks(VANETs),high mobility of vehicular environment and frequently changed network topology call for a low delay end-to-end routing algorithm.In this paper,we propose a Multi-Agent Reinforcement Learning(MARL)based decentralized routing scheme,where the inherent similarity between the routing problem in VANET and the MARL problem is exploited.The proposed routing scheme models the interaction between vehicles and the environment as a multi-agent problem in which each vehicle autonomously establishes the communication channel with a neighbor device regardless of the global information.Simulation performed in the 3GPP Manhattan mobility model demonstrates that our proposed decentralized routing algorithm achieves less than 45.8 ms average latency and high stability of 0.05%averaging failure rate with varying vehicle capacities.
基金supported by the National Natural Science Foundation of China(No.61901403)the Science and Technology Key Project of Fujian Province,China(Nos.2021HZ021004 and 2019HZ020009)+3 种基金the Open Research Fund of National Mobile Communications Research Laboratory,Southeast University(No.2023D10)the Youth Innovation Fund of Natural Science Foundation of Xiamen(No.3502Z20206039)the Science and Technology Key Project of Xiamen(No.3502Z20221027)the Xiamen Special Fund for Marine and Fishery Development(No.21CZB011HJ02).
文摘It has always been difficult to achieve accurate information of the channel for underwater acoustic communications because of the severe underwater propagation conditions,including frequency-selective property,high relative mobility,long propagation latency,and intensive ambient noise,etc.To this end,a deep unfolding neural network based approach is proposed,in which multiple layers of the network mimic the iterations of the classical iterative sparse approximation algorithm to extract the inherent sparse features of the channel by exploiting deep learning,and a scheme based on the Sparsity-Aware DNN(SA-DNN)for UAC estimation is proposed to improve the estimation accuracy.Moreover,we propose a Denoising Sparsity-Aware DNN(DeSA-DNN)based enhanced method that integrates a denoising CNN module in the sparsity-aware deep network,so that the degradation brought by intensive ambient noise could be eliminated and the estimation accuracy can be further improved.Simulation results demonstrate that the performance of the proposed schemes is superior to the state-of-the-art compressed sensing based and iterative sparse recovery schems in the aspects of channel recovery precision,pilot overhead,and robustness,particularly under unideal circumstances of intensive ambient noise or inadequate measurement pilots.
基金supported by the National Key Research and Development Program of China (No. 2018YFB1003605)Foundations of CARCH (No. CARCH201704)+3 种基金the National Natural Science Foundation of China (No. 61472312)Foundations of Shaanxi Province and Xi’an ScienceTechnology Plan (Nos. B018230008 and BD34017020001)the Foundations of Xidian University (No. JBZ171002)
文摘Despite the rapid development of mobile and embedded hardware, directly executing computationexpensive and storage-intensive deep learning algorithms on these devices’ local side remains constrained for sensory data analysis. In this paper, we first summarize the layer compression techniques for the state-of-theart deep learning model from three categories: weight factorization and pruning, convolution decomposition, and special layer architecture designing. For each category of layer compression techniques, we quantify their storage and computation tunable by layer compression techniques and discuss their practical challenges and possible improvements. Then, we implement Android projects using TensorFlow Mobile to test these 10 compression methods and compare their practical performances in terms of accuracy, parameter size, intermediate feature size,computation, processing latency, and energy consumption. To further discuss their advantages and bottlenecks,we test their performance over four standard recognition tasks on six resource-constrained Android smartphones.Finally, we survey two types of run-time Neural Network(NN) compression techniques which are orthogonal with the layer compression techniques, run-time resource management and cost optimization with special NN architecture,which are orthogonal with the layer compression techniques.
基金partially supported by the National Natural Science Foundation of China (Nos. 61190110, 61272456, and 61472312)the open fund ITDU14004/KX142600011+1 种基金supported by the overall innovation project of Shaanxi Province Science and Technology Plan (No. 2012KTZD02-03-03)the Fundamental Research Funds for the Central Universities (Nos. JB151002, K5051323005, and BDY041409)
文摘Lack of physical activity is becoming a killer of our healthy life. As a solution for this negative impact,we propose Smart Care to help users to set up a healthy physical activity habit. Smart Care can monitor a user's activities over a long time, and then provide activity quality assessment and suggestion. Smart Care consists of three parts, activity recognition, energy saving, and health feedback. Activity recognition can recognize nine kinds of daily activities. A hybrid classifier that uses less power and memory with satisfactory accuracy was designed and implemented by utilizing the periodicity of target activity. In addition, a learning-based energy saver was introduced to reduce energy consumption by adjusting sampling rates and the set of features adaptively. Based on the type and duration of the activity recorded, health feedback in terms of the calorie burned was given. The system could provide quantitative activity quality assessment and recommend future physical activity plans. Through extensive real-life testing, the system is shown to achieve an average recognition accuracy of 98.0% with a minimized energy expenditure.
基金financially supported by the Central University Special Fund Basic Research and Operating Expenses(No.GK201702005)the Natural Science Foundation of Shaanxi Province,China(No.2017JM6091)the Fundamental Research Funds for the Central Universities(No.2017TS011)
文摘A single output Q-switched Nd:GdVO4 laser with a reflective graphene oxide(GO) saturable absorber was demonstrated. The shortest pulse duration in the Q-switched laser is 115 ns, and the output power ranges from1.23 W at 1.71 MHz to 2.11 W at 2.50 MHz when the pump power rises from 7.40 to 10.90 W with the utilization of GO Langmuir–Blodgett(LB) films based on the convenient and low-cost LB technique. To the best of our knowledge, it is the highest output power in a Q-switched laser with a GO saturable absorber.
基金partially supported by the National Natural Science Foundation of China(No.41171323)Jiangsu Provincial Natural Science Foundation(No.BK2012018)+2 种基金the Key Laboratory of Geo-Informatics of National Administration of Surveying,Mapping and Geoinformation of China(No.201109)partially supported by the Priority Academic Program Development of Jiangsu Higher Education Institutions(PAPD)the Fundamental Research Funds for the Central Universities.
文摘Conventional change detection approaches are mainly based on per-pixel processing,which ignore the sub-pixel spectral variation resulted from spectral mixture.Especially for medium-resolution remote sensing images used in urban landcover change monitoring,land use/cover components within a single pixel are usually complicated and heterogeneous due to the limitation of the spatial resolution.Thus,traditional hard detection methods based on pure pixel assumption may lead to a high level of omission and commission errors inevitably,degrading the overall accuracy of change detection.In order to address this issue and find a possible way to exploit the spectral variation in a sub-pixel level,a novel change detection scheme is designed based on the spectral mixture analysis and decision-level fusion.Nonlinear spectral mixture model is selected for spectral unmixing,and change detection is implemented in a sub-pixel level by investigating the inner-pixel subtle changes and combining multiple composition evidences.The proposed method is tested on multi-temporal Landsat Thematic Mapper and China–Brazil Earth Resources Satellite remote sensing images for the land-cover change detection over urban areas.The effectiveness of the proposed approach is confirmed in terms of several accuracy indices in contrast with two pixel-based change detection methods(i.e.change vector analysis and principal component analysis-based method).In particular,the proposed sub-pixel change detection approach not only provides the binary change information,but also obtains the characterization about change direction and intensity,which greatly extends the semantic meaning of the detected change targets.
基金Central University Special Fund Basic Research and Operating Expenses(GK201702005)Natural Science Foundation of Shaanxi Province(2017JM6091)+1 种基金National Natural Science Foundation of China(NSFC)(61705183)Fundamental Research Funds for the Central Universities(2017TS011)
文摘In this article, we report on an experimentally generated soliton and bound-state soliton passively mode-locked erbium-doped fiber laser by incorporating a saturable absorber(SA) made of MoS2∕fluorine mica(FM) that was fabricated with the Langmuir–Blodgett(LB) method. The FM substrate is 20 μm thick and easy to bend or cut,like a polymer. However, it has a higher damage threshold and a better thermal dissipation than polymers.In addition, the LB method can be used to fabricate a thin film with good uniformity. In this study, the modulation depth, saturable intensity, and unsaturated loss of the SA are measured as 5.9%, 57.69 MW∕cm2, and13.4%, respectively. Based on the SA, a soliton mode-locked laser is achieved. The pulse duration, repetition rate, and signal-to-noise ratio are 581 fs, 15.67 MHz, and 65 dB, respectively. By adjusting the polarization controller and pump power, we obtain a bound-state soliton mode-locked pulse. The temporal interval between the two solitons forming the bound-state pulse is 2.7 ps. The repetition rate of the bound-state pulses is proportional to the pump power. The maximum repetition rate is 517 MHz, corresponding to the 33 rd harmonic of the fundamental repetition rate. The results indicate that the MoS2∕FM LB film absorber is a promising photonic device in ultrafast fiber lasers.
基金funded by the National Natural Science Foundation of China(Nos.21776184 and 21476147).
文摘The dissolution of collagen in ionic liquids(ILs)was highly dependent on the polarity of ILs,which was influenced by their sorts and concentrations.Herein,the solubility and dispersion degree of collagen in two sorts of ILs,namely 1-ethyl-methylimidazolium tetrafluoroborate([EMIM][BF4])with low polarity and 1-ethyl-3-methylimidazolium acetate([EMIM][Ac])with high polarity in a concentration range from 10% to 70% at 10℃ were investigated.When 150 mg of collagen was added to 30 mg of ILs,the minimum soluble collagen concentration was 0.02 mg/mL in 70%[EMIM][BF4]with lowest polarity and the maximum was 3.57 mg/mL in 70%[EMIM][Ac]with highest polarity,which indicates that soluble collagen and insoluble collagen fibers were both present.For insoluble collagens,differential scanning calorimetry showed that the thermal-stability was weakened when increasing the ILs concentration and polarity,and the fiber arrangement was looser with a more uniform lyophilized structure,observed by atomic force microscopy and scanning electron microscopy.For soluble collagens,electrophoresis patterns and Fourier transform infrared spectroscopy showed that no polypeptide chain degradation occurred during dissolution,but the thermal denaturation temperature decreased by 0.26℃~7.63℃ with the increase of ILs concentrations,measured by ultra-sensitive differential scanning calorimetry.Moreover,the aggregation of collagen molecules was reduced when ILs polarity was increased as determined by fluorescence measurements and dynamic light scattering,which resulted in an increased loose fiber arrangement observed by atomic force microscopy.If the structural integrity of collagen needs to be retained,then the ILs sorts and concentrations should be considered.
基金supported by the National Key RD Program of China (2019QY0600)the National Science Fund for Distinguished Young Scholars (62025205)+1 种基金the National Natural Science Foundation of China (Grant Nos.62032020,61960206008,62102317,62002292)the Natural Science Basic Research Plan in Shaanxi Province of China (2020JQ-207).
文摘1 Introduction Withthe rapid development of the E-commerce,more consumers turn to shop online.When online shoppers search for products using keywords,the related products appear with their brief descriptions.We conduct a real-world survey and find that most shoppers are dissatisfied with the existing"onefit-all product descriptions"and they have to spend more time to scan detail pages.However,handcrafting the attractive product descriptions is always costly.
基金This work was supported by the National Natural Science Foundation of China(Nos.61671396 and 91638204).
文摘Estates,especially those of public securityrelated companies and institutes,have to protect their privacy from adversary unmanned aerial vehicles(UAVs).In this paper,we propose a reinforcement learning-based control framework to prevent unauthorized UAVs from entering a target area in a dynamic game without being aware of the UAV attack model.This UAV control scheme enables a target estate to choose the optimal control policy,such as jamming the global positioning system signals,hacking,and laser shooting,to expel nearby UAVs.A deep reinforcement learning technique,called neural episodic control,is used to accelerate the learning speed to achieve the optimal UAV control policy,especially for estates with a large area,against complicated UAV attack policies.We analyze the computational complexity for the proposed UAV control scheme and provide its performance bound,including the risk level of the estate and its utility.Our simulation results show that the proposed scheme can reduce the risk level of the target estate and improve its utility against malicious UAVs compared with the selected benchmark scheme.
基金This work was supported by the National Natural Science Foundation of China(Nos.61971366,61671396,and 61901403)the Youth Innovation Fund of Xiamen(No.3502Z20206039)the Natural Science Foundation of Fujian Province of China(No.2020J01430).
文摘The video transmission in the Internet-of-Things(IoT)system must guarantee the video quality and reduce the packet loss rate and the delay with limited resources to satisfy the requirement of multimedia services.In this paper,we propose a reinforcement learning based energy-efficient IoT video transmission scheme that protects against interference,in which the base station controls the transmission action of the IoT device including the encoding rate,the modulation and coding scheme,and the transmit power.A reinforcement learning algorithm state-action-reward-state-action is applied to choose the transmission action based on the observed state(the queue length of the buffer,the channel gain,the previous bit error rate,and the previous packet loss rate)without knowledge of the transmission channel model at the transmitter and the receiver.We also propose a deep reinforcement learning based energy-efficient IoT video transmission scheme that uses a deep neural network to approximate Q value to further accelerate the learning process involved in choosing the optimal transmission action and improve the video transmission performance.Moreover,both the performance bounds of the proposed schemes and the computational complexity are theoretically derived.Simulation results show that the proposed schemes can increase the peak signal-to-noise ratio and decrease the packet loss rate,the delay,and the energy consumption relative to the benchmark scheme.