混杂系统的一致输入输出对状态稳定性

混杂系统的输入输出对状态稳定性是混杂控制系统领域极富挑战性的课题之一. 为了观测混杂系统的状态,本文提出了一类混杂系统的一致输入输出对状态稳定的充分条件,分析了混杂系统的一致输入输出对状态稳定性、光滑Lyapunov函数存在性和状态模估计器存在性三者之间的关系. 借助状态模估计器将混杂系统化为受扰动系统,获得了受扰动系统一致输入输出对状态稳定性的结果,并进一步证明了混杂系统的一致输入输出对状态稳定性.     

TL-SVM:一种迁移学习新算法

迁移学习旨在利用大量已标签源域数据解决相关但不相同的目标域问题. 当与某领域相关的新领域出现时, 若重新标注新领域, 则样本代价昂贵, 丢弃所有旧领域数据又十分浪费. 对此, 基于SVM算法提出一种新颖的迁移学习算法—–TL-SVM, 通过使用目标域少量已标签数据和大量相关领域的旧数据来为目标域构建一个高质量的分类模型, 该方法既继承了基于经验风险最小化最大间隔SVM的优点, 又弥补了传统SVM不能进行知识迁移的缺陷. 实验结果验证了该算法的有效性.     

基于多尺度耦合的密集残差网络红外图像增强

为了提升非制冷红外热像仪的图像质量,满足低对比度弱小区域的观瞄与锁定的需求,提出了一种基于多尺度密集残差网络的红外图像超分辨重建模型,该模型的基本框架是通过级联多个残差特征进行学习,以粗到细的方式重建高分辨率图像。首先提出一种多尺度跨域融合模块,通过对不同感受野的分支结果进行融合,不仅可以融合不同感受野的互补信息,还可有助于提升梯度收敛和特征传输;然后叠加多个跨域融合模块,并采用残差特征学习进行优化,最终学习出高分辨率细节信息。仿真实验结果表明,所提出的超分辨模型能够较好的超分辨重建效果,在微弱结构保持和点目标保持上的性能也更加突出。所提的模型已经在海思嵌入式深度学习平台上实现了高质量的红外增强,具有较高的工程应用价值。     

基于语境距离度量的拉普拉斯最大间距判别准则

线性拉普拉斯判别准则(Linear Laplacian discrimination, LLD)作为一种非线性特征提取方法得到了较为成功的运用. 然而通过分析得知在具体使用LLD方法的过程中还会面临小样本以及如何确定原始样本空间类型的问题. 因此, 本文引入语境距离度量并结合最大间距判别准则的基本原理提出一种基于语境距离度量的拉普拉斯最大间距判别准则(Contextual-distance metric based Laplacian maximum margin criterion, CLMMC). 该准则不但在一定程度上避免小样本问题, 而且由于语境距离度量更关注输入样本簇内在的本质结构而不是原始样本空间的类型, 从而降低了该准则对特定样本空间的依赖程度. 同时通过引入计算语境距离度量的新算法并结合QR分解的基本原理, 使得CLMMC在处理高维矢量模式数据时更具适应性和效率. 并从理论上讨论CLMMC准则具有的基本性质以及与LLD准则的内在联系. 实验证明CLMMC准则具有上述优势.     

Linear combination of densities and its direct estimation framework with applications

In this paper, typical learning task including data condensation, binary classification, identification of the independence between random variables and conditional density estimation is described from a unified perspective of a linear combination of densities, and accordingly a direct estimation framework based on a linear combination of Gaussian components (i.e., Gaussian basis functions) under integrated square error criterion is proposed to solve these learning tasks. The proposed direct estimation framework has three advantages. Firstly, different from most of the existing state-of-the-art methods in which estimating each component’s density in this linear combination of densities and then combining them linearly are required, it can directly estimate the linear combination of densities as a whole, and it has at least comparable to or even better approximation accuracy than the existing density estimation methods. Secondly, the time complexity of the proposed direct estimation framework is O(l ³) in which l is the number of Gaussian components in this framework which are generally viewed as the Gaussian distributions of the clusters in a dataset, and hence l is generally much less than the size of the dataset, so it is very suitable for large datasets. Thirdly, this proposed framework can be typically used to develop alternative approaches to classification, data condensation, identification of the independence between random variables, conditional density estimation and the similarity identification between multiple source domains and a target domain. Our preliminary results about experiments on several typical applications indicate the power of the proposed direct estimation framework.     

A new multi-objective bi-level programming model for energy and locality aware multi-job scheduling in cloud computing

How to reduce power consumption of data centers has received worldwide attention. By combining the energy-aware data placement policy and locality-aware multi-job scheduling scheme, we propose a new multi-objective bi-level programming model based on MapReduce to improve the energy efficiency of servers. First, the variation of energy consumption with the performance of servers is taken into account; second, data locality can be adjusted dynamically according to current network state; last but not least, considering that task-scheduling strategies depend directly on data placement policies, we formulate the problem as an integer bi-level programming model. In order to solve the model efficiently, specific-design encoding and decoding methods are introduced. Based on these, a new effective multi-objective genetic algorithm based on MOEA/D is proposed. As there are usually tens of thousands of tasks to be scheduled in the cloud, this is a large-scale optimization problem and a local search operator is designed to accelerate convergent speed of the proposed algorithm. Finally, numerical experiments indicate the effectiveness of the proposed model and algorithm.     

Analysis of aging effects - From transistor to system level

Due to shrinking feature sizes in integrated circuits, additional reliability effects have to be considered which influence the functionality of the system. These effects can either result from the manufacturing process or external influences during the lifetime such as radiation and temperature. Additionally, modern technology nodes are affected by time-dependent degradation i.e. aging.Due to the age-dependent degradation of a circuit, processes on the atomic scale of the semiconductor material lead to charges in the oxide silicon interface of CMOS devices, altering the performance parameters of the device and subsequently the behavior of the circuit. With the continuous downscaling of modern semiconductor technologies, the impact of these atomic scale processes affecting the overall system characteristics becomes more and more critical. Therefore, aging effects need to be assessed during the design phase and actions have to be taken guaranteeing the correct system functionality throughout a system's lifetime.This work presents methods to investigate the influence of age-dependent degradation as well as process-variability on different levels. An operating-point dependent sizing methodology based on the g_m/I_D-method extended to incorporate aging, which aims at developing aging-resistent circuits is presented. Additionally, the sensitivity of circuit performances in regard to aging can be determined. In order to investigate the reliability of a complex system on behavioral level, a modeling method to represent the performance of system components in dependence of aging and process variability is introduced.     

Microfluidic Surface‐Enhanced Raman Scattering Sensors Based on Nanopillar Forests Realized by an Oxygen‐Plasma‐Stripping‐of‐Photoresist Technique

A novel surface‐enhanced Raman scattering (SERS) sensor is developed for real‐time and highly repeatable detection of trace chemical and biological indicators. The sensor consists of a polydimethylsiloxane (PDMS) microchannel cap and a nanopillar forest‐based open SERS‐active substrate. The nanopillar forests are fabricated based on a new oxygen‐plasma‐stripping‐of‐photoresist technique. The enhancement factor (EF) of the SERS‐active substrate reaches 6.06 × 10⁶, and the EF of the SERS sensor is about 4 times lower due to the influence of the PDMS cap. However, the sensor shows much higher measurement repeatability than the open substrate, and it reduces the sample preparation time from several hours to a few minutes, which makes the device more reliable and facile for trace chemical and biological analysis.     

Fault-Tolerant Recursive Least-Squares Computations on a Mesh-Connected Parallel Processor

Recursive least squares (RLS) is a popular iterative method used for the modeling of systems while in operation. RLS provides an estimate for unknown parameters of a system based on some known parameters and inputs and outputs of that system. This technique is used frequently in digital signal processing and control applications, where it is not possible to completely determine the current state of the system. The RLS procedure incurs intensive computations in every iteration of the algorithm. To implement RLS in situ at a reasonable sampling rate, the complexity of the system's model must be reduced, or the available computing power must be increased. This paper examines methods for increasing the computing power by implementing RLS algorithms on a parallel processing platform. While there has been a large body of research on using parallel processors for the computation of adaptive algorithms, little of this research has examined fault tolerant aspects. As fault tolerance is a critical aspect of any real-time system, this work will examine some factors that should be considered when implementing a real-time adaptive algorithm on a parallel processor system.