多通道图注意力解耦社交推荐方法

提出具有解耦能力的多通道图注意力社交推荐模型,该模型主要包括深度聚类模块、多通道图注意力聚合模块和评分预测模块。其中,深度聚类模块用于对用户和项目进行分组,并利用聚类结果将用户社交图和用户项目图拆分成多个用户社交子图及用户项目子图,以学习用户兴趣分组及用户对不同类项目的兴趣；多通道图注意力聚合模块学习不同子图对预测结果的注意力；评分预测模块将学习到的用户表示向量和项目表示向量输入多层感知机进行评分预测。在多个真实数据集上的实验结果表明:提出的方法优于其他社交推荐算法。与最新的用于社交推荐的图神经网络方法相比,在Ciao和Epinions数据集上,均方根误差分别降低了2.26%和2.07%,平均绝对误差分别降低了2.58%和3.06%。     

Cost-based due-date assignment with the use of classical and neural-network approaches

Traditional methods of due-date assignment presented in the literature and used in practice generally assume cost-of-earliness and cost-of-tardiness functions that may bear little resemblance to true costs. For example, practitioners using ordinary least-squares (OLS) regression implicitly minimize a quadratic cost function symmetric about the due date, thereby assigning equal second-order costs to early completion and tardy behavior. In this article the consequences of such assumptions are pointed out, and a cost-based assignment scheme is suggested whereby the cost of early completion may differ in form and/or degree from the cost of tardiness. Two classical approaches (OLS regression and mathematical programming) as well as a neural-network methodology for solving this problem are developed and compared on three hypothetical shops using simulation techniques. It is found for the cases considered that: (a) implicitly ignoring cost-based assignments can be very costly; (b) simpler regression-based rules cited in the literature are very poor cost performers; (c) if the earliness and tardiness cost functions are both linear, linear programming and neural networks are the methodologies of choice; and (d) if the form of the earliness cost function differs from that of the tardiness cost function, neural networks are statistically superior performers. Finally, it is noted that neural networks can be used for a wide range of cost functions, whereas the other methodologies are significantly more restricted. © 1997 John Wiley & Sons, Inc.     

一种通用有效的神经网络映射算法

首先分析了神经网络映射的本质,神经网络拓扑结构和神经计算过程,在充分考虑负载均衡与通讯开销的基础上,引入了时间步的思想,提出了一种通用有效的神经网络映射算法,最后给出的对多种拓扑结构的神经网络的测试结果证明了该算法的通用性和有效性。     

面向射频指纹识别的高效IQ卷积网络结构

现有应用于射频指纹识别的卷积网络对时序同相正交(in-phase and quadrature-phase,IQ)信号的处理都是将其简单视为图像进行的,存在识别准确率低和计算量大的问题。针对以上问题,提出了一种基于IQ相关特征的卷积神经网络结构。该网络分步提取了IQ相关特征及时域特征,通过自适应平均池化获得了各通道特征均值,并用单个全连接层进行分类。实验结果表明,较传统卷积网络结构,所提网络在多种场景下的识别准确率更高,并且计算量更小。     

BP神经网络和D-S证据理论的目标识别

目标识别是指挥自动化系统的一个重要组成部分,针对现代战争对抗手段不断增强的特点,运用BP神经网络和D-S证据理论探索作战飞机机型的识别方法.前端采用3层BP神经网络结构,以传感器接收数据为输入,以神经网络输出作为证据,后端对不同传感器的证据按D-S理论进行融合,得到待识别目标的识别概率.经由MATLAB编程对国内外几种主要机型的识别进行仿真研究,与现行目标识别方法相比较,能够更快速、准确、可靠地识别飞机目标,较好地满足了空战中作战指挥系统对飞机机型识别的需求.     

基于Delphi-BP神经网络的装备保障能力评估

针对评估指标数量过多可能给评估过程及结果带来的弊端,在构建装备保障能力评估指标体系基础上,给出了基于Delphi法的指标体系筛选的方法步骤,通过计算累计贡献率,对指标体系进行了筛选,降低了评估模型的输入维度。建立了评估部队装备保障能力的三层BP神经网络模型,利用Matlab神经网络工具箱对网络进行了训练和仿真,结果显示误差小于10-3。最后,通过对比评估,验证了方法的有效性和正确性。     

基于深度森林的网络流量分类方法

随着网络应用的迅猛发展,流量分类在网络资源分配、流量调度和网络安全等诸多研究领域受到广泛关注。现有的机器学习流量分类方法对流量数据特征的选取和分布要求苛刻,导致在实际应用中的复杂流量场景下分类精确度和稳定度难以提高。为了解决样本特征属性的复杂性给分类性能带来的不利影响,引入了基于深度森林的流量分类方法。该算法通过级联森林和多粒度扫描机制,能够在样本数量规模和特征属性选取规模有限的情况下,有效地提高流量整体分类性能。通过网络流量公开数据集Moore对支持向量机、随机森林和深度森林机器学习算法进行训练和测试,结果表明基于深度森林的网络流量分类器的分类准确率能够达到96. 36%,性能优于其他机器学习模型。     

智能搜救机器人在障碍地形的自主构型规划

为了解决带有辅助摆臂的智能搜救机器人自动规划构型以实现自主越障的难题,提出一种能够适应复杂地面形状的搜救机器人越障构型规划新方法,其核心是一种高适应性、高效率的机器人姿态预测算法。通过将地形表示为离散的点集,建立了搜救机器人的单侧姿态预测数学模型;进一步提出了快速求解该问题的算法,每秒可预测1 000～1 500个姿态。基于此,设计了机器人越障过程中状态、动作的评价指标,运用动态规划算法与滚动优化思想构建了具有优化能力的、能够实时运行的构型规划器。仿真与实物实验的结果表明,该方法能够使机器人自主调整构型穿越复杂地形,且相较强化学习算法和人工操作具有更平稳的越障效果。     

神经网络在声纳目标识别中的研究进展

综述了神经网络在声纳信号预处理、目标特征提取及其自动分类方面的研究进展,简要分析了神经网络模型的特点和在应用中的局限性。并对该领域的发展趋势提出了自己的看法。     

A misclassification cost‐minimizing evolutionary–neural classification approach

Machine learning algorithms that incorporate misclassification costs have recently received considerable attention. In this paper, we use the principles of evolution to develop and test an evolutionary/genetic algorithm (GA)‐based neural approach that incorporates asymmetric Type I and Type II error costs. Using simulated, real‐world medical and financial data sets, we compare the results of the proposed approach with other statistical, mathematical, and machine learning approaches, which include statistical linear discriminant analysis, back‐propagation artificial neural network, integrated cost preference‐based linear mathematical programming‐based minimize squared deviations, linear integrated cost preference‐based GA, decision trees (C 5.0, and CART), and inexpensive classification with expensive tests algorithm. Our results indicate that the proposed approach incorporating asymmetric error costs results in equal or lower holdout sample misclassification cost when compared with the other statistical, mathematical, and machine learning misclassification cost‐minimizing approaches. © 2006 Wiley Periodicals, Inc. Naval Research Logistics, 2006.     

基于人工神经网络的操作手IKP 求解

人工神经网络以其信息的并行处理与分布式存储、自组织和自适应等特点,在机器人领域显示了极大的应用潜力。本文探讨人工神经网络在机器人操作手ＩＫＰ求解中的应用,试图求得操作手ＩＫＰ的一组或所有可行解。文章利用ＢＰ网络对操作手ＩＫＰ的求解进行了讨论和分析,并给出了具体的求解实例。     

Non‐greedy heuristics and augmented neural networks for the open‐shop scheduling problem

In this paper we propose some non‐greedy heuristics and develop an Augmented‐Neural‐Network (AugNN) formulation for solving the classical open‐shop scheduling problem (OSSP). AugNN is a neural network based meta‐heuristic approach that allows integration of domain‐specific knowledge. The OSSP is framed as a neural network with multiple layers of jobs and machines. Input, output and activation functions are designed to enforce the problem constraints and embed known heuristics to generate a good feasible solution fast. Suitable learning strategies are applied to obtain better neighborhood solutions iteratively. The new heuristics and the AugNN formulation are tested on several benchmark problem instances in the literature and on some new problem instances generated in this study. The results are very competitive with other meta‐heuristic approaches, both in terms of solution quality and computational times. © 2005 Wiley Periodicals, Inc. Naval Research Logistics, 2005.     

基于递归神经网络的飞机目标识别方法

针对宽带高距离分辨全极化雷达体制,提出了一种基于实时递归神经网络算法的飞机目标自动识别方法,实现了全极化下五类飞机目标的自动识别。实验结果表明,递归神经网络用于飞机目标识别是有效可行的。     

基于交叉熵和新转移函数的模糊神经网络分类器

针对目前普遍采用的误差平方和准则及Sigmoid转移函数在BP算法应用中存在的缺陷和不足,提出了基于交叉熵准则和新的S型转移函数构建的模糊神经网络分类器,并将这种分类器应用于心肌梗死的定位诊断,结果表明其训练效率和识别性能都明显优于传统的模糊神经网络。     

基于ANN技术的射击效率评定

将BP、RB、GRNN等人工神经网络引入火炮射击效率评定,其运算先确定相关参数、再根据毁伤概率产生输出数据、最后对数据进行检验,对训练好的网络进行仿真,并将计算与网络仿真的结果进行对比.结果表明,训练好的神经网络可较精确逼近解析解的结果,在计算量大的情况下该方法较解析计算更具优势.     

人工神经网络在管理信息系统中的应用

在传统管理信息系统的基础上,增加专家系统模块,即基于人工神经网络技术的预测分析模块,完成复杂的非线性预测以提高MIS智能化、自动化水平。该模块选用反向传输(BP:BackPropagation)神经网络模型来实现,通过网络的自适应学习和训练,找出输入和输出之间的内在联系,从而求得问题的解答。利用该专家系统对出版社图书市场销量进行预测。结果表明:该方法性能好,可作为预测图书售量的一种有效手段。     

基于广义回归神经网络的单位运营状况分类

单位的运营状况会直接影响股东和广大人民的利益,针对运营状况可以使用广义回归神经网络进行分类。由于广义回归神经网络中径向基函数的扩展参数Spread的选取会导致分类的准确率,提出了一种果蝇优化算法优化参数Spread的分类模型。充分利用了果蝇优化算法的寻优能力,将优化后的参数代入到广义回归神经网络中对单位的财务数据进行运营状况的分类。结果表明,与广义回归神经网络做比较,优化后的网络模型对数据的分类可以达到很高的准确率,在相关领域的分类上有非常大的实用性。     

Maneuvering target tracking of UAV based on MN-DDPG and transfer learning

Tracking maneuvering target in real time autonomously and accurately in an uncertain environment is one of the challenging missions for unmanned aerial vehicles(UAVs).In this paper,aiming to address the control problem of maneuvering target tracking and obstacle avoidance,an online path planning approach for UAV is developed based on deep reinforcement learning.Through end-to-end learning powered by neural networks,the proposed approach can achieve the perception of the environment and continuous motion output control.This proposed approach includes:(1)A deep deterministic policy gradient(DDPG)-based control framework to provide learning and autonomous decision-making capa-bility for UAVs;(2)An improved method named MN-DDPG for introducing a type of mixed noises to assist UAV with exploring stochastic strategies for online optimal planning;and(3)An algorithm of task-decomposition and pre-training for efficient transfer learning to improve the generalization capability of UAV's control model built based on MN-DDPG.The experimental simulation results have verified that the proposed approach can achieve good self-adaptive adjustment of UAV's flight attitude in the tasks of maneuvering target tracking with a significant improvement in generalization capability and training efficiency of UAV tracking controller in uncertain environments.     

基于学习矢量量化(LVQ)神经网络的雷达体制识别

由于现代战争中雷达体制的多样化、复杂化及其综合应用使得雷达体制识别中要处理大量复杂的高维数据,学习矢量量化(LVQ)神经网络不仅能处理有监督分类,而且相对于其他神经网络能以较小的计算量处理大量输入数据,所以采用LVQ对雷达体制进行识别,同时针对LVQ学习速率的变化可能引起学习算法不稳定,采用修正的学习速率算法.在简要介绍雷达体制和LVQ的基础上构造了LVQ神经网络对雷达体制进行分类.通过与径向基神经网络(RBFN)识别算法的仿真对比,证实了方法的有效性.     

矩阵变换特征与码序列联合学习的卷积码识别方法

现有基于深度学习的卷积码识别方法仍存在参数规模较大、识别性能较弱等不足。针对该问题,提出了一种基于矩阵变换特征与码序列联合学习的卷积码识别方法。将接收到的码字序列排列成矩阵形式,利用软信息剔除可靠性较低的码字,通过一种新的矩阵变换算法得到特征矩阵。在识别时,将原始码字矩阵和特征矩阵输入到具有多模态数据联合学习能力的网络模型,在神经网络中完成特征的提取融合与卷积码的识别。仿真结果表明,所提方法性能明显优于现有基于深度学习的识别方法,特别是对于高码率卷积码;当码率较低时,同样优于传统识别方法。当信噪比达到5 dB时,25种不同参数卷积码的识别率均可达到100%。