The building evacuation problem with shared information

In this article, the Building Evacuation Problem with Shared Information (BEPSI) is formulated as a mixed integer linear program, where the objective is to determine the set of routes along which to send evacuees (supply) from multiple locations throughout a building (sources) to the exits (sinks) such that the total time until all evacuees reach the exits is minimized. The formulation explicitly incorporates the constraints of shared information in providing online instructions to evacuees, ensuring that evacuees departing from an intermediate or source location at a mutual point in time receive common instructions. Arc travel time and capacity, as well as supply at the nodes, are permitted to vary with time and capacity is assumed to be recaptured over time. The BEPSI is shown to be NP‐hard. An exact technique based on Benders decomposition is proposed for its solution. Computational results from numerical experiments on a real‐world network representing a four‐story building are given. Results of experiments employing Benders cuts generated in solving a given problem instance as initial cuts in addressing an updated problem instance are also provided. © 2008 Wiley Periodicals, Inc. Naval Research Logistics, 2008     

超低照度下微光图像增强神经网络损失函数设计分析

超低照度下(环境照度小于2×10~(-3)lux)微光图像具有低信噪比、低对比度等特点,使目标难以辨识,严重影响观察效果。为了提高超低照度下微光图像质量,设计了一种用于微光图像增强的卷积自编码深度神经网络,并针对传统的均方误差损失函数不符合人类视觉感知特性等问题,结合现有的全参考图像质量评价指标,研究了包括感知损失在内的几种损失函数,并提出了一种新的可微分损失函数。实验结果表明,在网络结构不发生改变的情况下,所提损失函数具有更好的性能,在提高微光图像信噪比和对比度的同时,能够有效地增强图像内部细节信息。     

结合全局与局部信息的主动轮廓分割模型

针对传统的基于区域的主动轮廓模型在分割灰度不均匀图像和噪声图像存在效果不佳的问题,提出结合全局项与局部项的主动轮廓分割模型。全局项由CV(Chan-Vese)模型的保真项构成,局部项的构建考虑局部区域信息的同时引入反映图像灰度特性的局部熵信息。依据图像灰度的特点,选择合理的全局项和局部项参数,并加入正则项保证曲线在演化过程中保持平滑,保障分割结果的可靠性。通过变分水平集方法最小化能量泛函,依据梯度下降流迭代更新水平集,完成曲线演化。采用模拟图像和实际图像进行实验分析,结果表明,所提出的结合全局项和局部项的主动轮廓模型可以高效地分割噪声严重以及灰度分布不均匀的图像。     

Optimal switchover times between two activities utilizing the same resource

The “gold‐mining” decision problem is concerned with the efficient utilization of a delicate mining equipment working in a number of different mines. Richard Bellman was the first to consider this type of a problem. The solution found by Bellman for the finite‐horizon, continuous‐time version of the problem with two mines is not overly realistic since he assumed that fractional parts of the same mining equipment could be used in different mines and this fraction could change instantaneously. In this paper, we provide some extensions to this model in order to produce more operational and realistic solutions. Our first model is concerned with developing an operational policy where the equipment may be switched from one mine to the other at most once during a finite horizon. In the next extension we incorporate a cost component in the objective function and assume that the horizon length is not fixed but it is the second decision variable. Structural properties of the optimal solutions are obtained using nonlinear programming. Each model and its solution is illustrated with a numerical example. The models developed here may have potential applications in other areas including production of items requiring the same machine or choosing a sequence of activities requiring the same resource. © 2002 Wiley Periodicals, Inc. Naval Research Logistics 49: 186–203, 2002; DOI 10.1002/nav.10008     

概念层次网络中对偶性设计

首先对概念层次网络（ＨＮＣ）理论作简要说明，然后根据ＨＮＣ中层符号个性化设计理论给出对偶性设计方案，和偶性的具体表现，并给出对偶性概念具体表现的注解。     

一种BP神经网络算法的改进模型

首先对BP网络的结构和算法进行了分析,针对BP网络收敛速度慢,容易陷入局部极小等问题,提出了一种改进的BP网络模型,并对该模型算法进行了改进,通过激活函数的选择,网络的初始化,学习率的调整和训练样本数据的处理等方法,可实现加快网络的收敛速度,并且较好的解决局部最优问题.     

石河子移民文化对当地少数民族文化的影响

文化通性使不同民族文化之间的学习和交融成为可能，新疆石河子是以汉族移民为主体的移民地区，不同地域移民带来的文化互相碰撞交融，与兵团文化结合形成了独具特色的石河子多元一体地域文化。本文从经济生活、宗教信仰和民俗习惯、教育观和生育观几个方面来分析石河子特殊的移民文化对当地少数民族文化产生的影响。     

弹性BP神经网络消除轮速传感器误差方法的研究

汽车轮速是汽车运动状态参数的主要信息源,是控制系统的核心,其精度直接影响这些系统的性能.为了提高轮速的精度,降低传感器的研制成本,提出了一种基于弹性BP神经网络的误差分析方法消除轮速传感器误差.将改进的BP神经网络--弹性BP神经网络用于误差分析,并提出误差匹配的算法.理论和仿真结果表明,该方法使绝对误差达到2×10-4>rad,能够有效地消除传感器误差,提高轮速信号的精度.     

潜射弹道导弹出水姿态分析

分析了潜艇深度、纵倾角、横倾角、海面风速、海面波长、平均波高等因素对潜射导弹的偏航角、俯仰角、滚动角、偏航角速度、俯仰角速度、滚动角速度的影响。由于实验数据比较分散,不适合神经网络的训练,所以首先对实验数据模糊化,并用以对神经网络的训练。用训练好的神经网络分析了潜艇单因素变化的导弹出水姿态。     

气液两相管流流型识别理论研究进展

综述了气液两相流流型识别理论及方法的研究进展。首先探讨了流型划分问题，然后针对水平管路、垂直管路及摇摆管路介绍了两相流流型转换机理及转换的边界条件；对于流型的在线识别，介绍了基于波动理论、神经网络和图像处理的流型识别方法。最后提出了气液两相管流流型识别中亟待开展的理论研究方向。