A nonatomic‐game model for timing clearance sales under competition

We deal with dynamic revenue management (RM) under competition using the nonatomic‐game approach. Here, a continuum of heterogeneous sellers try to sell the same product over a given time horizon. Each seller can lower his price once at the time of his own choosing, and faces Poisson demand arrival with a rate that is the product of a price‐sensitive term and a market‐dependent term. Different types of sellers interact, and their respective prices help shape the overall market in which they operate, thereby influencing the behavior of all sellers. Using the infinite‐seller approximation, which deprives any individual seller of his influence over the entire market, we show the existence of a certain pattern of seller behaviors that collectively produce an environment to which the behavior pattern forms a best response. Such equilibrium behaviors point to the suitability of threshold‐like pricing policies. Our computational study yields insights to RM under competition, such as profound ways in which consumer and competitor types influence seller behaviors and market conditions. © 2014 Wiley Periodicals, Inc. Naval Research Logistics 61: 365–385, 2014     

A sequential dynamic pricing model and its applications

Consider a sequential dynamic pricing model where a seller sells a given stock to a random number of customers. Arriving one at a time, each customer will purchase one item if the product price is lower than her personal reservation price. The seller's objective is to post a potentially different price for each customer in order to maximize the expected total revenue. We formulate the seller's problem as a stochastic dynamic programming model, and develop an algorithm to compute the optimal policy. We then apply the results from this sequential dynamic pricing model to the case where customers arrive according to a continuous‐time point process. In particular, we derive tight bounds for the optimal expected revenue, and develop an asymptotically optimal heuristic policy. © 2004 Wiley Periodicals, Inc. Naval Research Logistics, 2004.     

基于微变焦法的超分辨率成像技术及实验研究

从空间域的角度分析了微变焦法的原理,微变焦过程可等效为改变每个光敏元所对应物辐射率分布区域的位置和大小,由此建立微变焦过程的空域模型为一个线性方程组,由该模型可描述观察图像和目标高分辨率图像之间的关系,并通过求解该方程组得到目标图像。最后做了面阵CCD的超分辨率成像实验,结果表明微变焦法可明显提高成像系统的分辨率。     

空间非平稳高斯噪声背景下混合信号的DOA估计

针对空间非平稳高斯噪声背景下混合信号的DOA估计问题,提出了基于四阶累积量对混合波源的DOA估计算法。该算法首先构造四阶累积量矩阵,利用类似ESPR IT方法估计出所有非相关信源的DOA,然后对四阶累积量进行修正空间差分平滑去相干,消除非相关源和空间非平稳高斯噪声的影响,从而实现了对相干源的DOA估计。此方法通过分别估计非相关和相关或者相干信号的DOA,能够以较少的阵元实现对多个信号DOA估计,最后仿真实验也验证该算法的有效性。     

基于尺度选择性的空间数据源选择与预取策略

合理高效的数据源选择策略是提高空间信息检索系统效率的重要因素之一。针对以往研究中对空间尺度语义关注程度的不足,提出一种结合空间尺度语义选择检索数据源的方法。该方法综合考虑人眼视觉特性,计算出参与分布式查询计划生成的数据源集合,过滤在尺度意义上对最终查询结果无影响的数据源,并通过将数据源的简化表达进行预取,进一步减少了查询的整体执行代价。实验结果表明,提出的方法在大尺度查询和小尺度查询时均具有良好的性能。     

数字波束形成技术(DBF)在雷达中的应用

数字波束形成技术是天线波束形成原理与数字信号处理技术相结合的产物,其广泛应用于阵列信号处理领域。由于电磁工作环境的恶化和大量射频干扰的存在,在极低的信干噪比(SINR)条件下进行目标检测和信息提取十分困难。对于阵列系统,往往采用自适应数字波束形成(ADBF)技术,来抑制强干扰和方向性干扰对有用信号的影响。介绍了数字波束形成器的基本原理及其DSP的实现结构。     

相关MIMO信道下一种特征波束成型方案的性能研究

提出了相关多输入多输出信道下的一种空时分组码、空分复用与波束成型相结合的新方案.该方案利用了发射分集和空分复用所带来的分集和复用增益,并基于接收机反馈的信道相关矩阵设计了一种波束成型器.仿真结果表明,这种方案的性能较传统的方案在高信噪比时能获得较大的改善.     

Vertically restrictive pricing in supply chains with price‐dependent demand

We examine the behavior of a manufacturer and a retailer in a decentralized supply chain under price‐dependent, stochastic demand. We model a retail fixed markup (RFM) policy, which can arise as a form of vertically restrictive pricing in a supply chain, and we examine its effect on supply chain performance. We prove the existence of the optimal pricing and replenishment policies when demand has a linear additive form and the distribution of the uncertainty component has a nondecreasing failure rate. We numerically compare the relative performance of RFM to a price‐only contract and we find that RFM results in greater profit for the supply chain than the price‐only contract in a variety of scenarios. We find that RFM can lead to Pareto‐improving solutions where both the supplier and the retailer earn more profit than under a price‐only contract. Finally, we compare RFM to a buyback contract and explore the implications of allowing the fixed markup parameter to be endogenous to the model. © 2006 Wiley Periodicals, Inc. Naval Research Logistics, 2006.     

中文分词与命名实体识别的联合学习

将卷积结构引入循环神经网络,从而构建卷积循环神经网络。以此为基础,研究构建了面向中文分词与实体识别联合学习的序列标注模型。该模型依托卷积循环神经网络构建特征编码层,实现中文字序列局部空间特征和长距离时序依赖特征的联合提取;依托改进的循环神经网络构建标签解码层,实现标签序列长距离时序依赖的有效建模;依托统一的分词与实体识别序列标注模式实现分词信息与实体信息的联合学习,避免传统流水线法的误差传播问题。在人民日报语料和微软标注语料上的实验结果显示,该框架较传统统计模型和神经网络模型有显著的性能提升,尤其是在识别字数较多的命名实体时,其效果明显优于其他方法。     

Patterns in spatial point locations: Local indicators of spatial association in a minefield with clutter

The problem of detecting a minefield in the presence of clutter can be abstracted to that of detecting a spatial pattern within a set of point locations. The point locations are superpositions of several patterns, one of which corresponds to mines. In contrast to previous articles that take a formal, model‐based approach, this article proposes a statistical methodology that is distinctly exploratory. Each point location is considered separately, and its contributions to a global measure of spatial distances between locations are featured. Different patterns and unusual points can be more easily identified on the new scale. Both minefield data and simulated point patterns demonstrate the power of the method. © 2001 John Wiley & Sons, Inc., Naval Research Logistics 48:333–347, 2001