A dynamic estimation of distribution algorithm is presented for the problem of selecting hand colors which is formulated as a dynamic multi-objective optimization model. The proposed algorithm takes full advantage of the correlation among the CbCr (Cb stands for chrominance blue and Cr stands for chrominance red) vectors of hand colors and their dynamic law, which is divided into the following two stages, with each having its unique algorithmic characteristic. In the first stage, an interval method is employed to obtain a part of hand CbCr vectors which are in a rectangular region. Then new candidates are achieved by sampling around the non-dominated solutions in the second stage to form an offspring population for the subsequent evolutions. The advantage of the proposed algorithm lies in that the obtained non-dominated solutions contain CbCr vectors of a hand region and the hand region is segmented accurately from its background region by the CbCr vectors. The proposed optimization model and algorithm are applied to actual segmentation of dynamic hand-region, and the proposed algorithm is compared with two existing algorithms. The experimental results demonstrate that the proposed algorithm is effective and superior to the compared counterparts.

This study aims to reduce errors made by medical personnel that are caused by poorly read information on ventilator interfaces by determining the factors that enhance the legibility of typographic information, including the font size, font type, and character case, and determining whether personal protective equipment (PPE) has any impact on legibility. The rate of accuracy, response time (for making decisions and carrying out actions), and cognitive processing/efficiency through eye movement indicators (diameter of pupil (an index of cognitive effort) and average fixation time (index of level of engagement)) are used to measure the legibility. It is found that the font type and case have a significant impact on the legibility of the information. The Arial font is more legible than the Times New Roman font, and lowercase letters are more legible than uppercase letters. The PPE has no significant influence on the legibility. The font size has a significant effect on the response time, but not on the rate of accuracy, average fixation time and diameter of the pupil. The response time is greatly reduced with a font size of 20 pt. Therefore, a ventilator interface with information parameters that use a 20 pt Arial font with lowercase letters can offer greater legibility. The experimental results provide guidance for designing the typographic information of ventilator interfaces.

In today's era, a large number of geographical images cannot be stored properly due to their small memory and large number of characteristics. As HDFS has its limitations in storing small files, and in order to cope with the storage and reading needs of a large number of geographical images, a method is proposed to classify small files by means of a deep learning classifier, merge the classified images to establish an index, upload the metadata generated by the merger to a Redis cache database, upload the merged files to HDFS, and at the same time subsequently implement the reading of the files through the index file The merged files are then uploaded to HDFS, and the files are subsequently read from the index file. The experimental results show that the architecture proposed in this paper has high efficiency in geographic image storage, improves the efficiency of storing and accessing massive small images, and has good results.

针对噪声环境和时变转速工况下行星齿轮故障识别率低的问题,提出一种基于堆叠消噪自动编码器(SDAE)和门控循环单元神经网络(GRUNN)的行星齿轮故障识别方法.构建基于SDAE和GRUNN的混合模型,处理前后关联的时序数据,自动地从含噪样本中提取鲁棒故障特征;将行星齿轮故障诊断的训练样本看作该混合模型的输入数据,采用Adam优化算法和dropout技术训练该混合模型,实现多参数的优化,防止过拟合现象的发生;根据训练后的混合模型,利用softmax分类器识别待诊样本中行星齿轮的状态.通过行星齿轮的故障识别实验验证该方法的有效性,实验结果表明该方法具有较强的抗噪能力和时变转速适应能力.

Field Programmable Gate Array (FPGA) platform has been a popular choice for deploying Convolution Neural Networks (CNNs) as a result of its high parallelism and low energy consumption. Due to the limited on-chip computation and storage resources, FPGA clusters are becoming promising candidates to improve CNN throughputs. In this paper, we first put forward strategies to optimize the inter-board resource allocation in FPGA clusters. Then we model the multi-board cluster problem based on dynamic programming to get the optimal topology of the FPGA clusters. Experimental results show that typical well-known CNNs with our proposed FPGA cluster topology obtains an average throughput 4.33X than single-board solutions and 1.87X than other state-of-the-art multi-board solutions.

Non-intrusive load monitoring (a.k.a. power disaggregation) refers to identifying and extracting the consumption patterns of individual appliances from the mains which records the whole-house energy consumption. Recently, deep learning has been shown to be a promising method to solve this problem and many approaches based on it have been proposed. In this paper, we propose a sequence-to-subsequence learning method, which makes a trade-off between traditional sequence-to-sequence and sequence-to-point method, to balance the convergence difficulty in deep neural networks and the amount of computation in the inference period. We build our model based on conditional generative adversarial network that helps us avoid designing the loss function manually. In addition, we apply U-Net and Instance Normalization techniques to our model and demonstrate their effectiveness. Evaluations are performed on real-world data sets and we achieve the state-of-the-art performance.

Field Programmable Gate Array (FPGA) platform has been a popular choice for deploying Convolutional Neural Networks (CNNs) as a result of its high parallelism and low energy consumption. Due to the limitation of on-chip resources on a single board, FPGA clusters become promising solutions to improve the throughput of CNNs. In this paper, we firstly put forward strategies to optimize the resource allocation intra and inter FPGA boards. Then we model the multi-board cluster problem and design algorithms based on knapsack problem and dynamic programming to calculate the optimal topology of the FPGA clusters. We also give a quantitative analysis of the inter-board data transmission bandwidth requirement. To make our design accommodate for more situations, we provide solutions for deploying fully connected layers and special convolution layers with large memory requirement. Experimental results show that typical well-known CNNs with the proposed topology of FPGA clusters could obtain a higher throughput per board than single-board solutions and other multi-board solutions.

为解决小样本和噪声干扰下滚动轴承剩余寿命(RUL)预测准确率低的问题,提出一种基于信息最小二乘生成对抗网络(information least squares generative adversarial network,InfoLSGAN)和行动者-评论家(actor-critic,AC)算法的滚动轴承剩余寿命预测方法.将堆叠降噪自动编码器、信息生成对抗网络和最小二乘生成对抗网络相结合,构建InfoLSGAN,自动地从噪声数据中提取可解释的鲁棒特征,解决梯度消失问题;采用基于AC的训练算法训练InfoLSGAN,减少训练时间,加快收敛速度;根据训练后的InfoLSGAN,利用softmax分类器预测测试样本中滚动轴承的剩余寿命.通过滚动轴承加速疲劳寿命试验验证该方法的有效性.试验结果证明,当信噪比等于0时,该方法对滚动轴承测试样本的寿命预测准确率至少提高了10％.在小样本情况下,滚动轴承剩余寿命预测的平均准确率达95.84％.

Static random access memory– (SRAM) based field programmable gate arrays (FPGAs) are currently facing challenges of limited capacity and high leakage power. To solve this problem, non-volatile memory (NVM) is proposed as the alternative to build non-volatile FPGAs (NVFPGAs). Even though the feasibility of NVFPGA has been confirmed, the utilization of multiple level cells (MLCs) has not been fully exploited yet. In this article, we study architecture of MLC-based NVFPGAs, and propose five cluster structures. To give detailed comparisons and extensive discussions, we conduct experiments for area, performance and leakage power evaluation. Based on explorations of the characteristics of MLC-based NVFPGAs, we further present MLC-aware timing-driven packing method to improve delay. In critical paths, our proposed method reduces the overhead of the additional delay in slow MLC cells. Experiments show that, compared to SRAM-based FPGAs, the proposed architecture with the proposed CAD flow can reduce the area, critical path delay and leakage power by 31%, 10%, and 95%, respectively.

Since static random access memory (SRAM)-based field-programmable gate array (FPGA) has limited density and comparatively high leakage power, researchers have proposed FPGA architectures based on emerging nonvolatile memories (NVMs) to satisfy the requirements of data-intensive and low-power applications. Among all components, block random access memory (BRAM) has the severest endurance problem in FPGA. Unluckily, traditional wear leveling (TWL) strategies cannot be directly applied to nonvolatile FPGA because it may induce large performance overhead. In this article, we propose performance-aware wear leveling schemes for nonvolatile FPGA to improve its lifetime. Two strategies pertaining to coarse-grained wear leveling (C-Pearl) and fine-grained wear leveling (F-Pearl) are developed to balance inter-BRAM and intra-BRAM writes. Procedures, including static analysis, wear leveling-guided placement, and reconfiguration are discussed. A supportive circuit design is proposed, too. The evaluation shows that C-Pearl and F-Pearl can achieve 34% and 46% higher lifetime improvement and simultaneously 8% and 11% lower performance overhead than TWL.

Owing to high transmission ratio and compact structure, planetary gearboxes are widely used in industrial applications. However, due to running under noisy environment and time-varying rotational speed conditions, the acquired vibration signals are nonstationary, which seriously degrade the performance of fault diagnosis methods of planetary gears. To address this problem, a fault diagnosis method of planetary gears using a stacked denoising autoencoder (SDAE) and a gated recurrent unit neural network (GRUNN) is proposed in this paper. First, a hybrid model based on SDAE and GRUNN is developed to remove noise components from input data and process pre- and post-correlation time-series data. The training samples for planetary gear fault diagnosis are regarded as the input data of the developed hybrid model to automatically extract robust fault features. Then, the training process of the developed hybrid model is presented. The Adam optimization algorithm is utilized to optimize the parameters, and the dropout technique is employed to prevent from overfitting. Finally, a softmax classifier is employed to identify planetary gear states in test samples. The effectiveness of the proposed method is validated through a fault diagnosis experiment of planetary gears. The experimental results show that the proposed method possesses strong anti-noise ability and adaptability to time-varying rotational speed.

草地是宁夏陆地生态系统的重要组成部分,估算其净初级生产力(NPP)对宁夏草地可持续利用与管理至关重要.采用MODIS数据和CASA模型对2000-2015年间宁夏草地生态系统NPP进行了估算,通过一元线性回归趋势分析、Hurst指数等方法研究草地NPP的时空变化规律及未来演变趋势,并分析草地NPP与气象因子的相关性.结果表明:(1)基于CASA模型的宁夏草地NPP模拟精度高,其估算值与实测多年草地NPP均值具有良好的线性关系(R=0.93,P＜0.01),与MOD17产品的草地NPP空间分布基本一致.(2)近16a宁夏草地年均NPP为148.28 g C m-2 a-1,且存在波动上升的趋势,其线性增长率为3.84 gC m-2 a-1(P＜0.01).(3)宁夏草地NPP整体处于上升趋势,草地NPP增长的草地面积达98％,且其增率自南向北递减;宁夏草地NPP的Hurst指数在0.27-0.81之间,均值为0.53,大部分草地的NPP变化趋势具有较强同向持续性.(4)在年时间尺度上,宁夏草地NPP主要受降水量的影响,与气温的相关性较弱;在月时间尺度上,生长季草地NPP与月总降水量的相关性高,且不存在时间滞后响应现象,而与月均温的响应则存在1个月的时间滞后性,宁夏大面积分布的干草原与荒漠草原NPP对气温响应滞后是导致这一现象发生的主要原因.

The traditional local connection scheduling algorithms may appear to block during task allocation process. This paper proposes a value-based local connection scheduling algorithm for distributed video transcoding system. In the proposed method, we first introduce the concept of value to video transcoding tasks by considering the user's click rate, video's expected load consumption and duration. Then the value of each video task can be calculated through its required load and user clicks on the unit load and unit duration. By sorting the priority of transcoding tasks according to their values, the proposed algorithm can preferentially transcode those videos with higher values, making efficient use of load resources. Further when the expected load exceeds the remaining resources of nodes, our method will temporarily skip this blocked task to find the next one with the expected load less than remaining resources, until there is no suitable task redistributed. The experiment results show that compared with typical local connection scheduling algorithm, our method can improve the transcoding efficiency by 3.26%, reducing the transcoding time by 2.70%.

The vegetation in desert/grassland biome transition zones is part of a fragile ecosystem that is sensitive to climate change. Thus, in recent decades, studying vegetation activity in desert/grassland biome transition zones has become important. Here, vegetation activity and the evolutionary tendencies of the temporal and spatial differentiation of the phenology of the desert/grassland biome transition zones were analyzed based on the Normalized Difference Vegetation Index (NDVI) of the third-generation Global Inventory Modeling and Mapping Studies (GIMMS3g) dataset. Additionally, the relationship between vegetation activity and climatic factors was analyzed based on NDVI and global meteorological reanalysis data. The results showed that the vegetation phenology of desert/grassland biome transition zones exhibits sharply contrasting characteristics between the Northern and Southern hemispheres, particularly when comparing differences before and after the breakpoint in global climate change (1998). The length of the growing season (LOS) of the Northern Hemisphere was shorter after 1998 than before it, and the integral of the growing season (IOS) of the NDVI decreased correspondingly. By contrast, the LOS in the Southern Hemisphere was longer, and after 1998, the IOS of the NDVI increased compared to its previous value. The vegetation activity trend and the fluctuation of the desert/grassland biome transition zones in the last 30 years can be divided into nine combined modes. However, these features also have an obvious turning point in 1998. The effects of evapotranspiration and precipitation on vegetation activity were most obvious, and these climatic factors drove the phenology changes in the different regions. Global warming limited the vegetation activity in low-latitude areas, but promoted it in middle-latitude areas.

Exact fingertip positions are of particular importance to the fingertip-based human–computer interaction. We build a multi-objective optimization model for the problem of fingertip localization, and present a method to solve the above model based on evolutionary algorithms. When building the model, we take the positions of a series of pixels as the decision variable, the shape of the hand-edge curve corresponding to each of the pixels as one objective function, and the distance between each of the pixels and the gravity center of the palm as the other objective function. In addition, based on the correlation among the positions of pixels of the fingertip regions, we present a multi-objective estimation of distribution algorithm to solve the model so as to obtain the best pixel set, thus gaining the fingertip positions. The experimental results demonstrate the effectiveness of the proposed model and algorithm.

Vegetation,an important component of terrestrial ecosystems,tells the story of the ecological response to climate change.In the context of climate change,vegetation changes have directly affected substance exchanges and energy balance of regional land-atmosphere interaction processes.Additionally,in China,the vegetation activity of varied terrestrial ecosystems has responded to global climate change in significantly different ways.As a vital remote sensing indicator,the normalized difference vegetation index (NDVI) reflects vegetation growth and coverage,and time series analysis can describe the temporal evolution and spatial variability of vegetation very well.Thus,studying the spatiotemporal variation of NDVI in terrestrial ecosystems not only reveals the response of vegetation in those ecosystems to climate change,but can provide a scientific basis for action on climate change and ecological considerations in an increasingly urbanized country like China.Based on the Global Inventory Modeling and Mapping Studies (GIMMS) NDVI3g dataset for 1982-2012,and the macro structure data of the Chinese terrestrial ecosystem,this study utilized one-dimensional linear regression,ensemble empirical mode decomposition (EEMD),and correlation analysis to examine spatiotemporal variation in the NDVI of each terrestrial ecosystem in China.It also analyzed the relationship between NDVI and climate events over a recent 30-year period.The results show that the average normalized difference vegetation index (ANDVI) and maximum normalized difference vegetation index (MNDVI) increased (0.0029/10 a,P＜0.05 and 0.0076/10 a,P＜0.01,respectively) over the same period.These phenomena indicate the activity of vegetation strengthened in China.In addition,owing to its sensitivity to water-thermal factors,MNDVI can be a useful supplement for monitoring vegetation change.Furthermore,the areas with significant increases in vegetation activities were mainly the farmland and forest ecosystems in the eastern monsoon region,and the areas with significant decreases in vegetation activities were mainly the forests in Northeast China and the desert ecosystem in Northwest China.With ANDVI and MNDVI showing linear trends at different significance levels,the NDVIs of the farmland,forest,grassland,and water and wetland ecosystems showed a general upward trend with large fluctuations,while the NDVI of the desert ecosystem decreased,with obvious vegetation degradation.However,EEMD detected an inflection point among the trends in the five ecosystems that was distinct from the linear trend.This occurred in the 1990s and early 21st century when the NDVI trend shifted from increasing to decreasing.The decomposed results indicated that the NDVI of the ecosystems had oscillation characteristics on an inter-annual (1.9-7.6 years) and an inter-decadal (10.7-22.2 years) scales,with the former being statistically more prominent.The increasing spatial heterogeneity of the NDVI of different terrestrial ecosystems also showed inter-annual periods of 2.1-2.7 years,with increasing regional differences in vegetation activity.Furthermore,the short period-pulsation of the air-sea system from the E1 Ni(n)o-Southern Oscillation and solar activity influenced vegetation activity via climatic factors,which was evidently related to the oscillation characteristics in all ecosystems.

Evapotranspiration (ET),a link between hydrological and ecological processes,is very important to the stability of terrestrial ecosystems.In arid and semi-arid regions,where more than 90 ％ of the effective precipitation is consumed by ET,variations in ET processes may weaken the stability of a grassland ecosystem plagued by water stress.Ningxia is located in the north China farming-grazing transitional zone;long term excessive conversion to farmland and overgrazing have greatly decreased the ecological function of the grassland ecosystem and generated serious changes in ET processes.Background information on grassland ET in the region is unknown,due to inadequate research on the characteristics of ET in undisturbed grassland in Ningxia.Research on the ET characteristics of grassland ecosystems within fragile ecological environments are important for regional agriculture and animal husbandry development,ecological restoration and reconstruction,and the construction of eco-barriers.This study,based on MOD16 ET data,corresponding climate data and grassland type,analyzed the spatial-temporal changes in ET in Ningxia from 2000 to 2014 and predict future trends.Regression analysis,correlation analysis and rescaled range analysis were employed.The results indicated that during a recent 15-year period (2000-2014),the ET of grassland increased (with fluctuations) and average ET was 228.03 mm per year.The upward trend was not significant (P＞0.05) because of regional differences in precipitation and heat,and the annual and monthly ET of various grassland types presented obvious heterogeneity.The annual average ET also had a strong spatial heterogeneity (135.84-732.12 mm),higher ET in the south and lower in the north,but the fluctuation was small.More specifically,grassland ET mainly presented a decreasing pattern from north to south in northern Ningxia,and increasing from north to south in central and southern Ningxia.However,the variation in ET is small and the change is not significant in most of the study area.In the future,the increasing trend of ET in the grasslands will probably become significant and only small parts of the grassland system will experience reduced ET.Precipitation was the main factor determining the distribution pattern and inter-annual fluctuation of ET in grassland in Ningxia.Nevertheless,the influence of land use change caused by human activities in ET cannot be ignored.

The annual vegetation phenology in the desert/grassland biome transition zone of Northern Hemisphere was deduced from the GIMMS NDVI 3g dataset and the spatio-temporal changes of phenology during 1982-2012 were studied. At the same time, the climatic driving factors of phenology change were also analyzed by coupling global meteorological reanalysis data. The results show that the vegetation phenology in each desert/grassland biome transition zone during 1982-1997 had different characteristics. However, the regular phenomenon of vegetation phenology changed after 1998. The end of growth season (EOS) in almost all desert/grassland biome transition zones of Northern Hemisphere became earlier during 1998-2012, which the rate is 0.41 d/a. The length of growth season (LOS) shortened in each region except for the Sahel, which the average shorten rate is 0.88 d/a. The correlation analysis between vegetation phenology and climatic factors show that the phenology of desert/grassland biome transition zone is affected by climate significantly and the main driving factors are different in each region. In the middle and high latitudes, the temperature is the key factor to limit the vegetation activity and the temperature rising can promote the start of growth season earlier. However, the precipitation increasing could weaken the growth of vegetation in these regions. In contrast, the water deficit that caused by the high temperature could shorten the vegetation growing season, and that is the key impact factor in the low latitudes of Northern Hemisphere. Finally, the relationship between phenology and climatic factors in different time-lags were also analyzed. The results show that the phenology change of desert/grassland biome transition zone is responded obviously to temperature change. In addition, there has a time lag effect in the response of phenology change to evapotranspiration. Nevertheless, this rule does not exist between phenology change and precipitation.