一种基于深度学习的视网膜病变图像识别方法

为解决医学上糖尿病性视网膜病变图像人工识别困难、精度差等问题,提出一种基于多特征融合的卷积神经网络识别方法。在VGG-16模型的基础上,通过融合每层网络上的局部特征,增强模型的特征提取能力。选用Softmax分类器,使病变图像识别更加准确。使用OpenCV图像处理工具采用加噪、上下左右不同角度翻转、调节对比度等5种方式扩充训练集。实验结果表明,基于多特征融合的深度学习框架图像识别系统在数据集上的平均识别精度达到94.23%,相较于Alex-Net、Google-Net、Compact-Net、ResNet-101等模型分别提高了10.56%、7.80%、6.01%、0.02%,验证了该方法的有效性。该模型具有很好的鲁棒性。     

基于改进型小数据量法的局域网流量预测

基于Takens理论对混沌时间序列进行相空间重构,对小数据量法进行如下改进:利用C-C算法计算嵌入维和延迟时间;以功率对频率加权并采用求平均的方法计算平均周期,使小数据量法更加完善。使用改进前、后的小数据量法分别仿真计算Lorenz系统混沌时间序列的Lyapunov指数并预测混沌时间序列,并计算实测局域网流量时间序列的最大Lyapunov指数并预测局域网流量时间序列。仿真及实验结果均表明,采用改进型小数据量法进行流量预测,精度更高、速度更快、预测点数更多。     

莫言小说扭曲性语言变奏浅析——透视“约定俗成”对语言审美灵性的遗忘与压抑

试从词法扭曲的扭曲方式入手,分析莫言小说的扭曲性语言构造及其相关特征,进一步揭示其使用后的超常表达效果,以此来透视语言本体论提出的"约定俗成性",对语言运用及审美过程中语言灵性的遗忘与压抑。     

中国翻译教学现状与复合型翻译人才培养

目前,在我国翻译教学火热开展的同时,也存在着很多问题.本文概括了我国翻译教学的现状,总结了其中存在的问题,并建议采用新型翻译人才培养模式,即复合型翻译人才培养道路.相信复合型翻译人才培养是解决翻译教学领域难题的有效途径.     

提高高等学校课堂教学有效性的应对策略

探讨高等学校课堂教学现状,分析存在的问题,找寻其中原因,提出应对策略：学生提高自我管理能力、转变学习方法、明确学习目的和增加专业认同感;教师端正教学态度、改变教学方法、提高教学能力;学校实行样本课程、完善教学基础设施、健全和改革学生和教师评价制度等.     

均匀线阵双基地MIMO雷达目标角度估计

在高斯色噪声背景下,提出了高阶累积量最小范数算法,实现了均匀线阵双基MIMO雷达的波离方向和波达方向的联合估计。最小范数算法利用双基MIMO雷达的联合导向矢量矩阵与噪声子空间正交,用噪声子空间的全部噪声奇异矢量构成最小范数矢量,相比MUSIC算法计算量小。仿真结果也表明在低信噪比时,最小范数算法的估计性能较好。     

多酸化合物的药学应用进展

介绍了多金属氧酸盐化学的发展过程,综述了多酸化合物的药学应用进展,包括其在抗艾滋病（HIV）、抗肿瘤、抗病毒等方面的研究情况.同时,展望了未来多酸药学应用研究的主要领域及亟需解决的关键问题.     

Yang–Baxter equations and quantum entanglements

In this paper some results associated with a new type of Yang–Baxter equation (YBE) are reviewed. The braiding matrix of Kauffman–Lomonaco has been extended to the solution (called type-II) of Yang–Baxter equation (YBE) and the related chain Hamiltonian is given. The Lorentz additivity for spectral parameters is found, rather than the Galilean rule for the familiar solutions (called type-I) of YBE associated with the usually exact solvable models. Based on the topological basis, the N-dimensional solution of YBE is found to be the Wigner D-functions. The explicit examples for spin-\(\frac{1}{2}\) and spin-1 have been shown. The extremes of \(\ell _1\)-norm of \(D\)-functions are introduced to distinguish the type-I from type-II of braiding matrices that also correspond to those of von Neumann entropy for quantum information.     

Protection of quantum correlations against decoherence

The protection of different quantum correlations, such as Bell nonlocality, quantum discord, and geometric quantum discord as trace distance against noise, is explored. By weak measurement and quantum measurement reversal, we show that the mentioned quantum correlations can be effectively preserved probabilistically from the decoherence due to amplitude damping. The results will play an important role in the experiments using the quantum correlations as resource.     

Robust preparation of four-qubit decoherence-free states for superconducting quantum interference devices against collective amplitude damping

Based on the quantum Zeno dynamics, we present an approach for deterministic preparation of arbitrary four-qubit decoherence-free state of superconducting quantum interference devices with respective to collective amplitude damping in a decoherence-free way, namely, not only the form of the target state is free of decoherence, but also the whole process for preparation. The operation is fast and convenient since we only need to manipulate three weak laser pulses sequentially. Other decoherence effects such as cavity decay and the spontaneous emission of qubits are also taken into account in virtue of master equation, and the strictly numerical simulation signifies the final fidelity is high corresponding to the current experimental technology.