基于改进MobileFaceNet的人脸识别方法北大核心

针对目前大多数人脸识别算法参数多、计算量大,难以部署到移动端和嵌入式设备中的问题,提出了一种基于改进MobileFaceNet的人脸识别方法。通过对MobileFaceNet模型结构的调整,将bottleneck模块优化为sandglass模块,改良深度卷积和逐点卷积的相对位置,适当增大sandglass模块的输出通道数,从而减少特征压缩时的信息丢失,增强人脸空间特征的提取。实验结果表明:改进后的方法在LFW测试数据集上准确率达99.15%,模型大小和计算量分别仅为原算法的61%和45%,验证了所提方法的有效性。     

Well‐Combined Magnetically Separable Hybrid Cobalt Ferrite/Nitrogen‐Doped Graphene as Efficient Catalyst with Superior Performance for Oxygen Reduction Reaction

Catalysts with low‐cost, high activity and stability toward oxygen reduction reaction (ORR) are extremely desirable, but its development still remains a great challenge. Here, a novel magnetically separable hybrid of multimetal oxide, cobalt ferrite (CoFe₂O₄), anchored on nitrogen‐doped reduced graphene oxide (CoFe₂O₄/NG) is prepared via a facile solvothermal method followed by calcination at 500 °C. The structure of CoFe₂O₄/NG and the interaction of both components are analyzed by several techniques. The possible formation of Co/Fe N interaction in the CoFe₂O₄/NG catalyst is found. As a result, the well‐combination of CoFe₂O₄ nanoparticles with NG and its improved crystallinity lead to a synergistic and efficient catalyst with high performance to ORR through a four‐electron‐transfer process in alkaline medium. The CoFe₂O₄/NG exhibits particularly comparable catalytic activity as commercial Pt/C catalyst, and superior stability against methanol oxidation and CO poisoning. Meanwhile, it has been proved that both nitrogen doping and the spinel structure of CoFe₂O₄ can have a significant contribution to the catalytic activity by contrast experiments. Multimetal oxide hybrid demonstrates better catalysis to ORR than a single metal oxide hybrid. All results make the low‐cost and magnetically separable CoFe₂O₄/NG a promising alternative for costly platinum‐based ORR catalyst in fuel cells and metal‐air batteries.     

一类新的能够渐进达到Gilbert-Varshamov界的Alternant子类码

本文基于Maximum Distance Separable(MDS)码的Hamming重量分布提出一类新的二元Alternant子类码.分析表明这类新的子类码包含整个BCH码类,并且可以渐进达到Gilbert-Varshamov(GV)界.     

基于Gabor特征和字典学习的高斯混合稀疏表示图像识别

为了克服图像识别中光照,姿态等变化带来的识别困难,同时提高稀疏表示图像识别的鲁棒性,本文提出了一种基于Gabor特征和字典学习的高斯混合稀疏表示图像识别算法.高斯混合稀疏表示是基于最大似然估计准则,将稀疏保真度表示为余项的最大似然函数,最终识别问题转化为求解加权范数的优化逼近问题.本文算法首先提取图像的Gabor特征;然后对Gabor特征集进行字典学习,由于在学习过程中引入了Fisher准则作为约束,学习得到具有类别标签的新字典;最后使用高斯混合稀疏表示识别方法进行分类识别.在3个公开数据库(人脸数据库AR库和FERET库以及USPS手写数字库)上的实验结果验证了该算法的有效性和鲁棒性.     

A projection method for the numerical solution of linear systems in separable stiff differential equations

This paper deals with the efficient implementation of implicit methods for solving stiff ODEs, in the case of Jacobians with separable sets of eigenvalues. For solving the linear systems required we propose a method which is particularly suitable when the large eigenvalues of the Jacobian matrix are few and well separated from the small ones. It is based on a combination of an initial iterative procedure, which reduces the components of the vector error along to the nondominant directions of J and a projection Krylov method which reduces the components of the vector error along to the directions corresponding to the large eigenvalues. The technique solves accurately and cheaply the linear systems in the Newton's method, and computes the number of stiff eigenvalues of J when this information is not explicitly available. Numerical results are given as well as comparisons with the LSODE code.     

Probabilistic fault diagnosis method based on the combination of nest-loop fisher discriminant analysis and analysis of relative changes

Bias of data location and increase in data variations are two typical disturbances, which in general, simultaneously exist in the fault process. Targeting their different characteristics, a nested-loop fisher discriminant analysis (NeLFDA) algorithm and relative changes (RC) algorithm are effectively combined for analyzing the fault characteristics. First, a prejudgment strategy is developed to evaluate the fault types and determine what changes are covered in the fault process. Two statistical indexes are defined, which conduct Monte Carlo based center fluctuation analysis and dissimilarity analysis respectively. Second, for the fault data containing those two faults simultaneously, a combined NeLFDA-RC algorithm is proposed for fault deviations modeling, which is termed as CNR-FD. Fault directions concerning bias of data location are extracted by the NeLFDA algorithm and then corresponding fault deviations are removed from the fault data. Then RC algorithm is performed on these fault data to extract directions concerning increase of data variations. These fault directions are used as reconstruction models to characterize each fault class. Particularly, the compromise between these two algorithms is determined by the Monte Carlo based center fluctuation analysis. For online applications, a probabilistic fault diagnosis strategy based on Bayes’ rule is performed to identify fault cause by discovering the right reconstruction models that can make the reconstructed monitoring statistics have the largest probabilities of belonging to normal condition. The motivation of the proposed algorithm is illustrated by a numerical case and the performance of the reconstruction models and the probabilistic fault diagnosis strategy are illustrated using pre-programmed faults from the Tennessee Eastman benchmark process and the real industrial process data from the cut-made process of cigarettes in some cigarette factory.     

Single‐layer perceptron and dynamic neuron implementing linearly non‐separable Boolean functions

This paper presents a single‐layer perceptron (SLP) scheme with an impulse activation function (IAF) and a dynamic neuron (DN) with a trapezoidal activation function (TAF). Combining with some interesting properties of the offset levels, it is shown that many linearly non‐separable Boolean functions can be realized by using only one SLPwIAF or one DNwTAF. In the present work, a few appropriate IAF and TAF are adopted, and the inverse offset level method is used for the design of the SLPwIAF synaptic weights and the DNwTAF templates. The XOR and NXOR Boolean operations with two inputs and all 152 non‐separable Boolean functions with three inputs can be easily implemented by one SLPwIAF or one DNwTAF. Finally, the entire set of 152 DNwTAF templates associated with 152 non‐separable Boolean functions of three inputs is completely listed. Copyright © 2008 John Wiley & Sons, Ltd.     

Realization of Multidimensional Digital Transfer Functions with Separable Numerator or Denominator Polynomials

In this paper a straightforward z-domain procedure for realizing multidimensional transfer functions with either numerator or denominator polynomial separable is presented. This procedure yields structures with minimum number of multipliers. It is shown that the number of delay elements can be reduced by selecting the optimum values for certain parameters. To facilitate this theorems that are applicable for some special cases are presented.     

On multiobjective combinatorial optimization and dynamic interim hedging of efficient portfolios

A dynamic portfolio policy is one that periodically rebalances an optimally diversified portfolio to account for time‐varying correlations. In order to sustain target‐level Sharpe performance ratios between rebalancing points, the efficient portfolio must be hedged with an optimal number of contingent claim contracts. This research presents a mixed‐integer nonlinear goal program (MINLGP) that is directed to solve the hierarchical multiple goal portfolio optimization model when the decision maker is faced with a binary hedging decision between portfolio rebalance periods. The MINLGP applied to this problem is formed by extending the separable programming foundation of a lexicographic nonlinear goal program (NLGP) to include branch‐and‐bound constraints. We establish the economic efficiency of applying this normative approach to dynamic portfolio rebalancing by comparing the risk‐adjusted performance measures of a hedged optimal portfolio to those of a naively diversified portfolio. We find that a hedged equally weighted small portfolio and a hedged efficiently diversified small portfolio perform similarly when comparing risk‐adjusted return metrics. However, when percentile risk measures are used to measure performance, the hedged optimally diversified portfolio clearly produces less expected catastrophic loss than does its nonhedged and naively diversified counterpart.     

转子故障特征数据分类的KPCA-BFDA方法

对非线性转子系统故障特征数据的分类方法进行了研究.在提出一种偏费歇判别分析法(biased fisher discriminatory analysis,简称BFDA)的前提下,进一步提出将核主成分分析法(kernel principal component analysis,简称KPCA)与偏费歇判别分析法相结合的数据集降维方法,该方法中的核主成分分析步骤用于构造剔除数据集冗余信息的降维数据集,偏费歇判别分析步骤用于进一步降低数据集维数并提高不同类别数据子集间的分离程度.对实例数据与典型故障数据的分类结果表明,提出的偏费歇判别分析法在具备费歇判别分析降维可分性能的基础上,具有更低计算复杂度的特点.提出的核主成分分析结合偏费歇判别分析的算法,其对应的降维结果能直接应用于线性分类器,且取得了较好的分类效果.