神经网络架构搜索研究进展与展望

神经网络架构搜索旨在针对不同任务,自动化地搜索得到性能最优的神经网络结构,是深度学习、计算机视觉技术结合当前现实需求应运而生的一大重要科学问题。对近年来神经网络架构搜索研究进行梳理、归类和评述;阐述神经网络架构搜索的定义和意义,全方位剖析当前研究所面临的难点与挑战;以此为基础,对主流的搜索策略进行阐述和归纳;探讨研究潜在的问题及未来颇具潜力的研究方向,以期推动该领域的进一步发展。     

基于累积前景理论的时间矩阵多属性综合评判模型

针对动态多属性方案排序优选决策问题,提出了一种基于累积前景理论的时间矩阵序列动态多属性综合评判系统模型。依据累积前景理论,将时间矩阵序列动态多属性决策转化为关于属性期望的益损时间矩阵序列动态决策,综合考虑各属性权重、各时间点的权重,计算各方案的最终综合前景值,并依据其大小对方案进行综合评判排序。最后给出了一个决策实例。     

动态矩阵控制在固体氧化物燃料电池中的应用

提出将动态矩阵控制(DMC)算法用于固体氧化物燃料电池(SOFC)的控制。为此,在分析SOFC动态机理模型的基础上,建立了其阶跃响应模型。控制目标是使SOFC的输出电压始终保持在一定范围内,并收敛到恒定的数值。在实际应用中,这样的算法应能提高SOFC输出功率的跟踪性能。仿真结果验证了这种方法的有效性。     

基于最小隶属度偏差的作战通道模糊优选

作战通道的优选对于充分发挥武器装备的效能,提高水面舰艇的作战能力具有十分重要的意义.通过分析作战通道效能评估的指标因素,确定指标相对优属度,应用最小隶属度偏差法,建立了作战通道的模糊优选模型,求解后得出了在该原则下的通道评估结论,为作战通道的优选提供了一种比较科学的定量分析方法.最后通过实例仿真计算得出了多条作战通道的优劣排序,为海战指挥员的科学决策提供参考.     

基于Vague集模糊值线性序法的雷达抗干扰效能评估

针对已有评估方法存在的不足,提出了一种基于Vague集模糊值线性序法的雷达抗干扰效能评估方法。分析了影响雷达抗干扰效能的指标因素,给出了各评价指标的模糊值表示,建立了基于Vague集模糊值线性序法的雷达抗干扰效能综合评价模型,得到了评价其效能的排序方法。最后通过实例分析验证,该方法评估结果准确,具有一定参考价值。     

基于估计协方差MME检测的频谱感知算法

针对小采样数据长度下,采样协方差矩阵对统计协方差矩阵估计不准,影响传统最大最小特征值(MME)检测算法检测性能的问题,提出一种基于逼近收缩(OAS)矩阵估计的改进MME检测算法。首先利用OAS估计量对采样数据做协方差矩阵估计,再对估计协方差矩阵特征值分解,将最大最小特征值之比作为检测统计量,克服了传统MME算法检测门限随采样点大幅波动的缺陷,提高了检测门限的鲁棒性。仿真结果表明,所提算法的检测门限具有鲁棒性,检测性能提高了1 d B~2 d B。     

防空导弹制导雷达综合抗干扰能力评估

分析了地空导弹武器系统制导雷达面临的主要干扰及其所采取的主要抗干扰措施,一般层次分析法(AHP)在评估某些模糊概念时存在缺陷。基于模糊层次分析法(FAHP),构造了模糊一致判断矩阵,建立了防空导弹制导雷达综合抗干扰能力评估模型,该模型具有快速、易于操作等优点,解决了在评价制导雷达综合抗干扰能力时难以定量分析的问题。     

指派问题的降阶优化算法

指派问题是运筹学中特殊线性规划中的一类问题。在现实生活中,指派问题非常普遍,常常可以见到各种各样的指派问题。通过对指派问题的数学模型进行分析,提出了与以往方法不同的求解指派问题的一种新的思路,通过对几个定理的研究,给出了一种新的求解方法——降阶优化算法。对求解指派问题提供了一种新的途径,在运筹学等领域有着较好的应用前景。     

Repairable consecutive‐k‐out‐of‐n: F system with Markov dependence

In this article, a model for a repairable consecutive‐k‐out‐of‐n: F system with Markov dependence is studied. A binary vector is used to represent the system state. The failure rate of a component in the system depends on the state of the preceding component. The failure risk of a system state is then introduced. On the basis of the failure risk, a priority repair rule is adopted. Then the transition density matrix can be determined, and the analysis of the system reliability can be conducted accordingly. One example each of a linear and a circular system is then studied in detail to explain the model and methodology developed in this paper. © 2000 John Wiley & Sons, Inc. Naval Research Logistics 47: 18–39, 2000     

Effect of tool rotational speed on the particle distribution in friction stir welding of AA6092/17.5 SiCp-T6 composite plates and its consequences on the mechanical property of the joint

This study investigates the effect of tool rotational speed (TRS) on particle distribution in nugget zone (NZ) through quantitative approach and its consequences on the mechanical property of friction stir welded joints of AA6092/17.5 SiC_p-T6 composite. 6 mm thick plates are welded at a constant tool tilt angle of 2° and tool traverse speed of 1 mm/s by varying the TRS at 1000 rpm, 1500 rpm and 2000 rpm with a taper pin profiled tool. Microstructure analysis shows large quantity of uniformly shaped smaller size SiC particle with lower average particle area which are homogeneously distributed in the NZ. The fragmentation of bigger size particles has been observed because of abrading action of the hard tool and resulting shearing effect and severe stress generation due to the rotation of tool. The particles occupy maximum area in the matrix compared to that of the base material (BM) due to the redistribution of broken particles as an effect of TRS. The migration of particles towards the TMAZ-NZ transition zone has been also encountered at higher TRS (2000 rpm). The microhardness analysis depicts variation in average hardness from top to bottom of the NZ, minimum for 1500 rpm and maximum for 2000 rpm. The impact strength at 1000 rpm and 1500 rpm remains close to that of BM (21.6 J) while 2000 rpm shows the accountable reduction. The maximum joint efficiency has been achieved at 1500 rpm (84%) and minimum at 1000 rpm (68%) under tensile loading. Fractographic analysis shows mixed mode of failure for BM, 1000 rpm and 1500 rpm, whereas 2000 rpm shows the brittle mode of failure.