一种鲁棒性的少样本学习方法

少样本学习是目前机器学习研究领域的一个热点,它能在少量的标记样本中学习到较好的分类模型.但是,在噪声的不确定环境中,传统的少样本学习模型泛化能力弱.针对这一问题,提出一种鲁棒性的少样本学习方法RFSL(Robust Few-Shot Learning).首先,使用核密度估计(Kernel Density Estimation,KDE)和图像滤波(Image Filtering)方法在训练集中加入不同的随机噪声,形成多个不同噪声下的训练集,并分别生成支持集和查询集.其次,利用关系网络的关系模块通过训练集端到端地学习多个基分类器.最后,采用投票的方式对各基分类器的最末Sigmoid层非线性分类结果进行融合.实验结果表明,RFSL模型可促进小样本学习快速收敛,同时,与R-Net以及其他主流少样本学习方法相比,RFSL具有更高的分类准确率,更强的鲁棒性.     

多重不确定的密码体制研究

针对现有密码体制中只有密钥不确定的特点,文章提出了多重不确定密码体制的概念,指出了可能实现的一些途径。文章分析了其优势,指出这类密码体制在大大增强安全性的同时,其可能带来的运算量的增加是可以忽略的。     

Decay rate of critical fluctuations in steam and in dilute steam-NaCl mixtures

The decay rate of critical fluctuations in steam and in a steam-NaCl mixture has been investigated experimentally with the aid of photon correlation spectroscopy. For pure steam, the measurements have been performed along seven isochores [(¦– _c¦)/_c<0.09] as a function of the temperatureT for (T–T_t)<1 K. The results have been compared with the values predicted by the renormalization-group theory written as a modification of the classical mode coupling theory. The agreement between experiment and theory is satisfactory along the critical isochore, but larger deviations are noted for _c when approching the transition temperatureT _t. The decay rate of a 0.1% (molar) dilute mixture of NaCl in H₂O has been measured along some near-critical isochores as a function of temperature. Its behavior, which is very different from that observed for pure steam, is dicussed.Paper presented at the Tenth Symposium on Thermophysical Properties, June 20–23, 1988, Gaithersburg, Maryland, U.S.A.     

Parallel heap: An optimal parallel priority queue

We describe a new parallel data structure, namely parallel heap, for exclusive-read exclusive-write parallel random access machines. To our knowledge, it is the first such data structure to efficiently implement a truly parallel priority queue based on a heap structure. Employing p processors, the parallel heap allows deletions of (p) highest priority items and insertions of (p) new items, each in O(log n) time, where n is the size of the parallel heap. Furthermore, it can efficiently utilize processors in the range 1 through n.This work was supported by U.S. Army's PM-TRADE contract N61339-88-g-0002, Florida High Technology and Industry grant 11-28-716, and Georgia State University's internal research support during spring and summer quarters, 1991.     

基于图像内容层次表征的遥感图像 分割方法

提出基于图像内容层次表征的高分辨率遥感图像快速多精度分割方法。首先根据初始分割结果建立区域邻接图(RAG),并将其定义为马尔可夫随机场(MRF);然后引入光谱、形状和边缘等图像特征进行层次合并,通过记录层次合并过程获得图像内容的层次表征;最后根据层次表征中不同层级对象之间的关系快速生成任意不同精度的分割结果,以满足不同应用的需求。利用QuickBird卫星图像进行实验和评价的结果表明,本文方法具有较高的精度和效率。     

改进FCM在交互式图像分割中的应用

为解决由于自然纹理的干扰而导致的分割图像边缘模糊问题,对模糊C均值聚类算法进行改进并应用于交互式图像分割中。用户通过输入种子点来获得目标和背景的主要特征,并将输入的种子点作为聚类中心点;提出全局空间相似性度量标准并引入Gabor能量滤波器来计算图像中各点到聚类中心的距离;算法首次引入边缘密度概念定义权重因子,根据图像特点,自适应地计算图像中任意一点的纹理特征和颜色特征在特征空间中所占比例,使得到的特征更加准确地描述图像的本质属性。对具有自然纹理背景的图像进行仿真实验,应用两种性能指标来比较本文所提算法与随机游走算法的分割精度。实验结果表明,本文算法分割精度高于模糊聚类和随机游走算法。     

Water table fluctuation with a random initial condition

The nonlinear Boussinesq equation is used to understand water table fluctuations in various ditch drainage problems. An approximate solution of this equation with a random initial condition and deterministic boundary conditions, recharge rate and aquifer parameters has been developed to predict a transient water table in a ditch-drainage system. The effects of uncertainty in the initial condition on the water table are illustrated with the help of a synthetic example. These results would find applications in ditch-drainage design.Notation A / tanh t - a lower value of the random variable representing the initial water table height at the mid point - a+b Upper value of the random variable representing the initial water table height at the midpoint - B tanh t - C 4/ - h variable water table height - h mean of the variable water table height - h _m variable water table height at the mid point - h _m mean of the variable water table height at the mid point - K hydraulic conductivity - L half spacing between the ditches - m ₀ initial water table height at the mid point - N Uniform rate of recharge - S specific yield - t time of observation - x distance measured from the ditch boundary - (4/SL)(NK)^1/2 - (L/4)(N/K)^1/2 - dummy integral variable     

基于K-距离拓扑的分布式数据存储方法

针对分布式数据存储算法通常需要较长的等待时间、且对海量数据加密时严重浪费计算资源和时间的问题,提出了一种基于K-距离拓扑的分布式数据存储方法.通过寻找K-距离拓扑子图来实现数据的安全放置,优先选择存取速度更快的节点和自身保护能力强的节点实现总体性能的提升.在Internet 2拓扑图与随机拓扑图下的仿真测试结果表明,所提出的方法能在满足安全距离约束的条件下选择到最优的数据存储节点,从而减小数据存取时间.     

基于3D scSE-UNet的肝脏CT图像半监督学习分割方法

针对分割神经网络需要大量的高质量标签但较难获取的问题,提出基于3D scSE-UNet的半监督学习分割方法. 该方法使用自训练的半监督学习框架,将包含改进的并行空间/特征通道压缩和激励模块（scSE-block+）的3D scSE-UNet作为分割网络. scSE-block+可以从图像空间和特征通道2个方面自动学习图像的有效特征,抑制无用冗余特征,更好地保留图像边缘信息. 在自训练过程中加入全连接条件随机场,对分割网络产生的伪标签进行边缘细化,提升伪标签的精确度. 在LiTS17 Challenge和SLIVER07数据集上验证所提出方法的有效性. 当有标签图像占训练集总图像的30%时,所提方法的Dice相似系数(dice score)为0.941. 结果表明,所提出的半监督学习分割方法可以在仅使用少量标注数据的情况下,取得与全监督分割方法相当的分割效果,有效减轻肝脏CT图像分割对专家标注数据的依赖.     

电能变换领域中随机脉宽调制技术研究概况及展望

相较于传统的脉宽调制技术,随机脉宽调制具备有效的谐波频谱分散能力,从而实现机电负载噪声的削弱。随机脉宽调制技术包括起步的随机开关调制、后续的随机脉冲位置与随机开关频率等3个基本方法。随着电力电子技术的整体发展,随机脉宽调制已从传统的载波调制逐渐拓展到空间矢量调制之中,应用范围日趋广泛,谐波频谱分散与功率谱峰值抑制的效果不断提升。同时,新的优化方案不断被提出,对谐波频谱及功率谱的分散效果亦在不断提高。梳理了随机脉宽调制的研究历程,总结了数种典型的随机脉宽调制的基本原理,并对未来随机脉宽调制的研究方向作出展望,为随机脉宽调制技术的深入研究提供参考。