Measuring Mental Workload Using ERPs Based on FIR,ICA, and MARA

Mental workload is considered to be strongly linked to human performance, and the ability to measure it accurately is key for balancing human health and work. In this study, brain signals were elicited by mental arithmetic tasks of varying difficulty to stimulate different levels of mental workload. In addition, a finite impulse response (FIR) filter, independent component analysis (ICA), and multiple artifact rejection algorithms (MARAs) were used to filter event-related potentials (ERPs). Then, the data consisting of ERPs, subjective ratings of mental workload, and task performance, were analyzed through the use of variance and Spearman’s correlation during a simulated computer task. We found that participants responded faster and performed better in the easy task condition, followed by the medium and high-difficulty conditions, which verifies the validity of the ERP filtering. Moreover, larger P2 and P3 waveforms were evoked as the task difficulty increased, and a higher task difficulty elicited a more enhanced N300. Correlation analysis revealed a negative relationship between the amplitude of P3 and the subjective ratings, and a positive relationship between the P3 amplitude and accuracy. The results presented in this paper demonstrate that a combination of FIR, ICA, and MARA methods can filter ERPs in the non-invasive real-time measurement of workload. Additionally, frontocentral P2, N3, and parietal P3 components showed differences between genders. The proposed measurement of mental workload can be useful for real-time identification of mental states and can be applied to human–computer interaction in the future.     

基于循环生成对抗网络的含遮挡人脸识别

人脸图像的遮挡会严重影响人脸识别准确率,当前处理带遮挡人脸识别的方法主要有丢弃法和修复法两种。丢弃法因忽略或丢弃大量遮挡区域的有效特征易造成识别准确率不高,而当前大多数修复法需要原图的相关信息,限制了其应用。针对现有含遮挡人脸识别方法存在的问题,提出一种基于循环生成对抗网络的人脸识别算法,通过利用2对生成器和判别器进行循环训练,实现遮挡人脸图像的盲修复,此过程不需要除遮挡区域外的原图信息。在此基础上,采用ResNet-50网络对修复后人脸进行识别,该网络通过跳跃连接消除深层卷积神经网络中的网络退化问题,能够降低网络训练的难度,且不会增加额外的参数和计算量。针对盲修复后人脸特征存在类内差异大和类间差异小的特性,引入一种能够量化类间距离的分类损失函数RegularFace作为识别网络损失函数。实验结果表明,与DCGAN+CNN算法相比,所提算法对不同遮挡类型和遮挡区域的人脸图像识别准确率均有所提高,当线性遮挡面积为40%时,所提算法的识别准确率提高了14.4个百分点。     

融合自动权重学习的深度子空间聚类

子空间聚类算法是一种面向高维数据的聚类方法,具有独特的数据自表示方式和较高的聚类精度。传统子空间聚类算法聚焦于对输入数据构建最优相似图再进行分割,导致聚类效果高度依赖于相似图学习。自适应近邻聚类（CAN）算法改进了相似图学习过程,根据数据间的距离自适应地分配最优邻居以构建相似图和聚类结构。然而,现有CAN算法在进行高维数据非线性聚类时,难以很好地捕获局部数据结构,从而导致聚类准确性及算法泛化能力有限。提出一种融合自动权重学习与结构化信息的深度子空间聚类算法。通过自编码器将数据映射到非线性潜在空间并降维,自适应地赋予潜在特征不同的权重从而处理噪声特征,最小化自编码器的重构误差以保留数据的局部结构信息。通过CAN方法学习相似图,在潜在表示下迭代地增强各特征间的相关性,从而保留数据的全局结构信息。实验结果表明,在ORL、COIL-20、UMIST数据集上该算法的准确率分别达到0.780 1、0.874 3、0.742 1,聚类性能优于LRR、LRSC、SSC、KSSC等算法。     

基于神威·太湖之光的非结构网格计算加速算法

在国产异构众核平台神威·太湖之光上的非结构网格计算具有稀疏存储、离散访存、数据依赖等特点,严重制约了众核处理器的性能发挥。为解决稀疏存储和离散访存问题,提出一种N阶对角染色算法,以有效平衡主从核计算并利用从核将全局访存转化为LDM访问。针对数据依赖造成的计算竞争问题,采用自适应和无依赖的任务划分方法,避免并行计算时的数据冲突。为对处理器架构和非结构网格计算进行优化,采用主核与从核异步并行的方式,差异化使用主从核以充分利用硬件资源,同时,取消处理器提供的寄存器通信机制,降低从核阵列的同步开销同时便于扩展到新一代神威平台。此外,使用计算访存异步重叠技术来充分隐藏访存延迟。利用SpMV、Integration、calcLudsFcc算子进行实验,结果表明,相比主核实现,组合加速算法在不同算例规模下平均取得了10倍的加速效果,加速比最高可达24倍,N阶对角染色算法相比非染色分块算法取得了超过5.8倍的性能加速,有效提升了数据局部性和计算并行度。该算法对有依赖关系的计算冲突算子同样具有良好的加速性能,验证了自适应和无依赖任务划分方法的有效性。     

Determinants of internal and external R&D offshoring: evidence from Spanish firms

This paper analyses the determinants of R&D offshoring of Spanish firms using information from the Panel of Technological Innovation. We find that being an exporter, international technological cooperation, continuous R&D engagement, applying for patents, being a foreign subsidiary, and firm size are factors that positively affect the decision to offshore R&D. In addition, we find that a lack of financing is an obstacle relatively more important for independent firms than for firms that belong to business groups. For these latter, we also obtain that the factors that influence the decision to offshore R&D differ depending on whether the firm purchases the R&D services within the group or through the market: a higher degree of importance assigned to internal sources of information for innovation as compared to market sources increases (decreases) the probability of R&D offshoring only through the group (market).     

Simultaneous Imaging of Endogenous Survivin mRNA and On‐Demand Drug Release in Live Cells by Using a Mesoporous Silica Nanoquencher

The design of multifunctional drug delivery systems capable of simultaneous target detection, imaging, and therapeutics in live mammalian cells is critical for biomedical research. In this study, by using mesoporous silica nanoparticles (MSNs) chemically modified with a small‐molecule dark quencher, followed by sequential drug encapsulation, MSN capping with a dye‐labeled antisense oligonucleotide, and bioorthogonal surface modification with cell‐penetrating poly(disulfide)s, the authors have successfully developed the first mesoporous silica nanoquencher (qMSN), characterized by high drug‐loading and endocytosis‐independent cell uptake, which is able to quantitatively image endogenous survivin mRNA and release the loaded drug in a manner that depends on the survivin expression level in tumor cells. The authors further show that this novel drug delivery system may be used to minimize potential cytotoxicity encountered by many existing small‐molecule drugs in cancer therapy.     

Probabilistic crack trajectory analysis by a dimension reduction method

This paper proposes a novel analysis method of stochastic crack trajectory based on a dimension reduction approach. The developed method allows efficiently estimating the statistical moments, probability density function and cumulative distribution function of the crack trajectory for cracked elastic structures considering the randomness of the loads, material properties and crack geometries. First, the traditional dimension reduction method is extended to calculate the first four moments of the crack trajectory, in which the responses are eigenvectors rather than scalars. Then the probability density function and cumulative distribution function of the crack trajectory can be obtained using the maximum entropy principle constrained by the calculated moments. Finally, the simulation of the crack propagation paths is realized by using the scaled boundary finite element method. The proposed method is well validated by four numerical examples performed on varied cracked structures. It is demonstrated that this method outperforms the Monte Carlo simulation in terms of computational efficiency, and in the meanwhile, it has an acceptable computational accuracy.     

Stochastic two-sided U-type assembly line balancing: a genetic algorithm approach

In this paper, a novel stochastic two-sided U-type assembly line balancing (STUALB) procedure, an algorithm based on the genetic algorithm and a heuristic priority rule-based procedure to solve STUALB problem are proposed. With this new proposed assembly line design, all advantages of both two-sided assembly lines and U-type assembly lines are combined. Due to the variability of the real-life conditions, stochastic task times are also considered in the study. The proposed approach aims to minimise the number of positions (i.e. the U-type assembly line length) as the primary objective and to minimise the number of stations (i.e. the number of operators) as a secondary objective for a given cycle time. An example problem is solved to illustrate the proposed approach. In order to evaluate the efficiency of the proposed algorithm, test problems taken from the literature are used. The experimental results show that the proposed approach performs well.     

Managing supply systems with partial information on shipment locations

This paper studies a supply system for a retailer who orders a single product from one manufacturer. Orders filled by the manufacturer pass through multiple transportation stages before reaching the retailer. Each stage represents either a physical location or a step in the delivery process. The lead time for a new order depends on the location of shipments against prior orders in transit. Shipments are not allowed to cross over in time. Thus, the movement of each shipment depends on the movements of shipments ahead of it and the resulting congestion. The retailer is able to track shipments as they move through the transportation channel. The retailer adopts an ordering policy that minimises the sum of his one-period holding and shortage costs, using available status information of shipments already in transit. The case where practical constraints prevent the retailer from obtaining a complete status of shipments at all stages in the transportation channel is considered. The methodology developed evaluates the value of partial shipment tracking information, and uses it to determine the optimal placement of a limited number of tracking devices. The methodology can also be used to evaluate the cost–benefit of placing additional tracking devices in the supply system.