The impact of reverse logistics process coordination on third party relationship quality: A moderated mediation model for multichannel retailers in the fashion industry

Multichannel retailing is a widely adopted strategy in the fashion industry. Companies in this industry find it a source of competitive advantage to invest in reverse logistics infrastructure. However, limited empirical studies investigate the enablers of the relationship quality between the retailers and their reverse logistics service providers. This study quantifies the impact of reverse logistics process coordination between retailers and logistics service providers on relationship quality. Moreover, it tests the mediating role of reverse logistics service quality and the moderating role of conflict frequency in this relationship. Data were collected through a survey using a purposive sample of 241 retail store managers from the fashion retail industry of Pakistan. For this purpose, a self-administered questionnaire was developed using a five-point Likert Scale to gauge the responses. Conditional process analysis was used to evaluate the moderated mediation model. The findings showed a significant positive impact of reverse logistics process coordination on the relationship quality with a logistics service provider, a significant positive mediation effect of reverse logistics service quality, and a significant moderation effect of conflict frequency on the indirect relationship. However, conflict frequency, contrary to the hypothesis in the study, strengthened the indirect relationship. Furthermore, the moderation effect of conflict frequency on the direct relationship was insignificant. This study will help managers better understand the best practices leading to effective management of reverse logistics processes, particularly product returns.     

X频段高频谱纯度频率合成器设计

针对高灵敏度接收机对频率合成器的高技术指标要求，构建了一种融合了直接模拟、直接数字以及间接数字的频率合成技术方案，根据该方案，成功实现了频率合成器的工程研制。通过测试，频率合成器相位噪声达-112 dBc/Hz@5 kHz，杂散抑制优于-75 dBc，频率分辨率小于1 kHz，10 MHz跳频时间约为13 μs，满足了高灵敏度接收机对频率合成器的高技术指标要求，为高纯度频率合成器的实现提供了一条新途径。     

基于MBP算法和深度学习的人脸识别

为了解决在深度学习提取人脸图像特征时,易忽略其局部结构特征和缺乏对其旋转不变性学习的问题,提出了一种基于单演局部二值模式(MBP)与深度学习相结合的高效率人脸识别方法。首先,用多尺度单演滤波器对图像进行滤波,得到幅值和方向信息;其次,用LBP算法和象限比特的方法进行编码,分块计算组合其直方图特征;然后,将提取的单演特征作为深度信念网络(DBN)的输入,逐层训练优化网络参数,得到优异的网络模型;最后,将训练好的DBN网络在ORL人脸数据库上进行人脸识别实验,进行识别率计算,其识别率为98.75%。所提出的方法使用无监督的贪婪算法,隐藏层设定为2层,使用反向传播算法优化网络。相较于已知的人脸识别方法,MBP+DBN算法对光照、表情和部分遮挡变化具有较好的鲁棒性,在人脸识别中识别率较高,具有一定的优势,为图像特征提供了一种新的识别方法。     

广义空间调制系统的正则化OMP检测算法

在广义空间调制（GSM）系统中，最大似然（ML）检测可以取得最优的检测性能，然而其计算复杂度随激活天线数的增加急剧增长。针对这一问题，提出了一种基于稀疏重构理论的低复杂度检测算法——正则化正交匹配追踪（ROMP）算法。该算法首先根据信道矩阵和当前残差的内积选取多个候选激活天线索引，接着对候选天线索引按正则化标准进行可靠性验证，剔除错误索引，缩小信号的搜索空间，最后通过求解最小二乘问题估计信号。仿真结果表明，与经典的正交匹配追踪（OMP）算法相比，所提算法以少许复杂度的增加为代价极大提升了检测性能，能够在检测性能与复杂度之间取得更好的折中。     

跳频信号的归一化多普勒频差最大似然估计

针对双站定位涉及的跳频信号多普勒频差估计问题，提出了归一化的频差最大似然估计算法，利用两个定位站接收到的跳频脉冲串信号，构建一个关于基准跳频频率多普勒频差的似然函数，通过网格搜索得到使似然函数最大的多普勒频差估计，既解决了跳频信号在不同跳频频率上多普勒频差不一致的问题，又充分利用在不同频率的脉冲串信号提高了多普勒频差估计精度。通过仿真对算法的性能进行了评估，结果表明，与基于子空间的算法比较，在脉冲数达到240个时，所提算法执行效率提升30%以上。     

Are REITs more resilient than non-REITs? Evidence from natural experiments

REITs draw attention from investors around the world, yet our understanding of the various risks associated with such securities is limited. Using the introduction of Arrowhead, a low-latency high-frequency trading platform, to the Tokyo Stock Exchange and the financial crisis of 2008 as natural experiments, we compare the resilience of REITs and equities in terms of liquidity and volatility. The results indicate that the introduction of Arrowhead improved the quality of the Japanese REIT market but also increased the probability of flash crashes. We also find that although the financial crisis significantly deteriorated overall equity market quality, the Japanese REIT market was resilient. Finally, using a difference-in-differences regression model, we show that the higher transparency and better price discovery of REITs, compared to non-REITS, protected them from the negative effects of the financial crisis and the introduction of Arrowhead. Overall, our analysis shows that REITs are more resilient than non-REITs.     

Market impact: a systematic study of the high frequency options market

This paper deals with a fundamental subject that has seldom been addressed in recent years, that of market impact in the options market. Our analysis is based on a proprietary database of metaorders—large orders that are split into smaller pieces before being sent to the market—on one of the main Asian markets. In line with our previous work on the equity market [Said, E., Bel Hadj Ayed, A., Husson, A. and Abergel, F., Market impact: A systematic study of limit orders. Mark. Microstruct. Liq., 2018, 3(3&4), 1850008.], we propose an algorithmic approach to identify metaorders, based on some implied volatility parameters, the at the money forward volatility and at the money forward skew. In both cases, we obtain results similar to the now well-understood equity market: Square-Root Law, Fair Pricing Condition and Market Impact Dynamics.     

基于改进指纹聚类的WLAN定位优化方法

将K-means聚类算法应用到无线局域网（WLAN）位置指纹定位中，虽然可以缩短定位时间，但是容易降低定位精度。为了解决此问题，提出了基于改进指纹聚类的WLAN定位优化方法。首先根据接收信号强度标准差来优化初始聚类中心的选取,然后对指纹数据进行聚类处理,最后进行在线定位。实验结果表明，与传统的WLAN位置指纹定位方法和K-means聚类定位方法相比，基于改进指纹聚类的定位优化方法不仅缩短了定位时间，还能有效提高定位精度。     

单频网多播系统中的开环半动态功率分配算法

在单频网多播传输中，传统的全反馈动态功率分配数算法需要根据每个时隙反馈的用户瞬时信道信息进行实时的调整，所以造成了资源分配频率快、上行反馈开销大的缺点。为了克服这个缺点，提出了一个低复杂度、没有用户反馈的单频网多播开环半动态功率分配算法。首先在各小区等功率分配的假设下，根据单频网的形状信息算出各小区等价信道增益，然后再根据这个增益值，实现满足速率需求情况下的各小区功率分配。仿真结果显示，与全反馈的动态功率分配算法相比，该算法以一小部分性能损失为代价，大大减少了单频网的上行反馈和资源分配的开销，因此更适用于实际的单频网多播系统。     

基于深度学习的跳频信号识别

针对跳频信号分选存在人工提取参数特征具有复杂性的问题，提出了一种基于深度学习的识别方法。首先对跳频信号进行短时傅里叶变换，得到二维的时频矩阵；接着提取信号的轮廓特征，构造三维矩阵作等高线图，并对等高线图进行预处理；最后把预处理后的等高线图输入到卷积神经网络中进行训练、测试，进而实现分类识别。仿真结果表明，在不需要复杂的人工提取参数特征的基础上，在分选率为100〖WT《Times New Roman》〗%〖WTBZ〗时，所提方法经裁剪处理下的信噪比为-15 dB，比支持向量机和传统K-Means聚类算法都低10 dB。实测数据的算法验证表明，所提方法能够将大疆精灵4Pro、hm无人机、司马航模X8HW以及大疆悟2这四类无人机正确分类。