Improved learning of I2C distance and accelerating the neighborhood search for image classification

Image-to-class (I2C) distance is a novel measure for image classification and has successfully handled datasets with large intra-class variances. However, due to the lack of a training phase, the performance of this distance is easily affected by irrelevant local features that may hurt the classification accuracy. Besides, the success of this I2C distance relies heavily on the large number of local features in the training set, which requires expensive computation cost for classifying test images. On the other hand, if there are small number of local features in the training set, it may result in poor performance.In this paper, we propose a distance learning method to improve the classification accuracy of this I2C distance as well as two strategies for accelerating its NN search. We first propose a large margin optimization framework to learn the I2C distance function, which is modeled as a weighted combination of the distance from every local feature in an image to its nearest-neighbor (NN) in a candidate class. We learn these weights associated with local features in the training set by constraining the optimization such that the I2C distance from image to its belonging class should be less than that to any other class. We evaluate the proposed method on several publicly available image datasets and show that the performance of I2C distance for classification can significantly be improved by learning a weighted I2C distance function. To improve the computation cost, we also propose two methods based on spatial division and hubness score to accelerate the NN search, which is able to largely reduce the on-line testing time while still preserving or even achieving a better classification accuracy.     

CCS-MAC: Exploiting the overheard data for compression in wireless sensor networks

Both the overhearing and overhearing avoidance in a densely distributed sensor network may inevitably incur considerable power consumption. In this paper we propose a so-called CCS-MAC (collaborative compression strategy-based MAC) MAC protocol which facilitates to exploit those overheard data that is treated useless in traditional MAC protocols for the purpose of cost and energy savings. Particularly the CCS-MAC enables different sensor nodes to perform data compression cooperatively with regard to those overheard data, so that the redundancy of data prepared for the link layer transmission can be totally eliminated at the earliest. The problem of collaborative compression is analyzed and discussed along with a corresponding linear programming model formulated. Based on it a heuristic node-selection algorithm with a time complexity of (O(N²)) is proposed to the solve the linear programming problem. The node-selection algorithm is implemented in CCS-MAC at each sensor node in a distributed manner. The experiment results verify that the proposed CCS-MAC scheme can achieve a significant energy savings so as to prolong the lifetime of the sensor networks so far.     

User-centric query refinement and processing using granularity-based strategies

Under the context of large-scale scientific literatures, this paper provides a user-centric approach for refining and processing incomplete or vague query based on cognitive- and granularity-based strategies. From the viewpoints of user interests retention and granular information processing, we examine various strategies for user-centric unification of search and reasoning. Inspired by the basic level for human problem-solving in cognitive science, we refine a query based on retained user interests. We bring the multi-level, multi-perspective strategies from human problem-solving to large-scale search and reasoning. The power/exponential law-based interests retention modeling, network statistics–based data selection, and ontology-supervised hierarchical reasoning are developed to implement these strategies. As an illustration, we investigate some case studies based on a large-scale scientific literature dataset, DBLP. The experimental results show that the proposed strategies are potentially effective.     

卫星链路上的SR协议性能分析

选择重传ARQ协议兼顾了停等ARQ协议可靠性高、退后N ARQ协议传输能力强、应用广泛的优势.传统的选择重传ARQ协议模型没有考虑窗口大小对协议性能的影响,导致选择重传ARQ协议模型分析能力不足.为此,通过对选择重传ARQ协议的运行进行分析,建立了与选择重传ARQ协议运行一致的数学模型,并通过仿真验证了所建模型的合理性...     

基于ATmega128的通信基站智能节能窗控制系统设计

介绍了一种采用ATmega128单片机设计的通信基站智能节能窗控制系统,通过传感器对通信基站的内外环境变量进行实时取样,让节能窗与空调联动工作,从而实现节能效果。实际应用表明,该系统在通信基站、机房等建筑内能明显减少空调运行时间,降低能耗。     

TD-LTE系统中RACH信号检测的仿真与实现

提出了TD-LTE系统中RACH(随机接入)信号检测算法.该算法利用ZC (Zadoff- Chu)序列和FFT(快速傅里叶变换)特点,结合循环相关函数为eNodeB提供快速有效的随机接入.仿真结果表明,在不同的信道环境下,该检测算法都能较好地工作,具有快速稳定、易于实现等优点.     

单级倒立摆的LQR控制和DMC控制Matlab仿真比较

建立了单级直线倒立摆的状态空间模型,分别从线性最优二次型（LQR）控制和模型预测控制的原理出发,在MAT-LAB/SIMULINK环境下,设计了LQR控制器和动态矩阵控制（DMC）控制器,对所建立的倒立摆模型对象进行系统控制仿真。仿真结果表明,作为先进控制的DMC控制的效果比作为最优控制的LQR控制效果更好,鲁棒性更强。     

一种双模式的运动估计算法

为了平衡运动估计中搜索算法的复杂度与搜索精度,基于双模式算法的思想,提出一种结合基于改进的粒子群算法(PSO)和十字搜索算法(ARPS)的双模式运动搜索算法.该算法对不同运动程度的图像采用不同的运动搜索算法(运动剧烈时采用PSO算法,运动平缓时采用ARPS算法),有效地结合了PSO的全局性特点以及ARPS的局部性特点,同时保持了ARPS的快速性.实验表明,该算法的整体性能高于传统的单模式运动估计算法以及已有的多模式运动估计算法.     

中位-四[二-(3,4-羧甲氧基)苯基]卟啉及其结合Re=O的初步研究

用密度泛函理论的B3LYP法,将中位-四[二-(3,4-羧甲氧基)苯基]卟啉及结合Re=O分子的几何构型优化,初步理论研究其前沿轨道和能量等。证明,与单独的卟吩环相比,T_(3,4)CPP的卟吩环平面性良好,环上的结构参数变化很小;Re=O与T_(3,4)CPP卟吩环的结合能明显大于与T_(3,4)CPP侧链羧基的结合能,且前者所形成配合物的稳定性显然高于后者,说明T_(3,4)CPP与~(188)Re=O结合的位点应在卟吩环上,而不在侧链羧基上。T_(3,4)CPP卟吩环结合Re=O而形成配合物的结构参数,与Vicente等Zn-卟啉衍生物的测定值相近。在卟吩环上配位时,T_(3,4)CPP卟吩环上的电荷重新分布,卟吩环碳原子的电子向4个氮原子转移:配位后,卟吩环的几何构型略收缩(0.14 A-0.19 A)和马鞍形扭转(5.7°)。     

2-(4-甲氧基苯基)亚肼基乙酸与乙醛酸反应密度泛函理论研究

本文应用量子化学密度泛函理论(DFT)研究了2-(4-甲氧基苯基)亚肼基乙酸与乙醛酸的反应机理及动力学行为,并研究了不同取代基对反应的影响.在RB3LYP/6-31G*水平上,优化了各反应通道的反应物、中间体、过渡态及产物的几何构型,计算了各驻点的振动频率、零点能和电荷分布.计算结果表明,该反应有2条反应通道,分别生成(2Z)和(2E)-3(4-甲氧基苯基二氮烯基)丙烯酸.产物发生了键长的平均化和电荷的重新分布.两反应通道具有相同的反应入口,Z式产物为主要产物,2-(4-甲氧基苯基)亚肼基乙酸中苯环对位被给电子基团取代,有利于反应进行.