ID	原文	译文
5844	为了提高稀疏最小二乘支持向量机对高维、异构数据的泛化性能,提出新型的基于压缩感知的稀疏多核最小二乘支持向量机算法。	In order to improve for high-dimensional sparse least squares support vector machine (SVM), the generalization performance of the heterogeneous data, a new method based on compression perception of sparse multi-core least squares support vector machine (SVM) algorithm.
5845	首先根据压缩感知理论,用正交匹配追踪算法对最小二乘支持向量机的支持向量进行稀疏化,再利用线性多核扩展法求出新的核函数矩阵。	First of all, based on compressed sensing theory, using the orthogonal matching pursuit algorithm for least squares support vector machine (SVM) to sparse support vector, linear multi-core extension method is a new kernel function matrix.
5846	将新的核矩阵应用到最小二乘支持向量机,得到稀疏多核最小二乘支持向量机的解,用稀疏的支持向量实现函数回归。	Apply new nuclear matrix to the least squares support vector machine (SVM), sparse multi-core least squares support vector machine solution, with sparse support vector to realize the function to return.
5847	理论分析与数据实验对比结果表明该模型对于高维、异构数据能够更快更准确地进行训练,大大提高了模型的泛化能力和运算速度。	Theoretical analysis and data contrast experiment results show that the proposed model for high-dimensional, heterogeneous data can more quickly and accurately for training, greatly enhance the generalization ability of the model and the computational speed.
5848	针对当前通信系统所采用的主要调制方式,提出了一种基于卷积神经网络和稀疏滤波的调制识别方法。	In view of the current communication system used by the main modulation method, this paper proposes a modulation recognition based on convolutional neural network and sparse filtering method.
5849	首先,分析了利用信号循环谱二维灰度图进行通信信号调制识别的可行性;	First of all, the paper analyzes the use of cyclic spectrum two-dimensional grayscale signal the feasibility of the communication signals modulation recognition;
5850	然后,通过降采样和裁剪技术对循环谱图预处理;最后,设计了深度卷积神经网络架构,并提出了稀疏滤波预训练的方法。	Then, by sampling and cutting technology of cyclic spectra pretreatment;Finally, the design depth of convolution neural network architecture, and puts forward the sparse filtering method in the process of the training.
5851	仿真结果表明:相比于经典的基于深度学习的调制识别方法,该方法模型简单,优化量少,且在小样本场景下性能最佳,具有很高应用价值。	Simulation results show that compared with the classical modulation recognition method based on the deep study, the method is simple in model, the optimal quantity is little, and the best performance under small sample scenario, has the very high application value.
5852	针对属性权重和专家权重全部未知的三角模糊数(triangular fuzzy number,TFN)多属性群决策问题,在TFN熵的基础上构造了确信度指标来量化对决策信息的信任程度,构建了TFN确信度(TFN certitude degree,TFNCD)算子,并证明了其置换不变性、幂等性和有界性等性质,结合支持度确定专家权重,提出了基于TFNCD算子的属性信息集结新方法。	For all unknown attribute weight and expert weight of triangular fuzzy number (triangular fuzzy number, TFN) multiple attribute group decision making problems, on the basis of TFN entropy construct confidence level indicators to quantify's trust in the decision-making information, constructs the TFN confidence level (TFN certitude degree, TFNCD) operator, and prove its displacement invariability, idempotence and boundedness nature, combined with support expert weight is determined, based on the attribute information gathering new TFNCD operator is proposed.
5853	最后,通过算例的对比分析验证了TFNCD算子及其集结方法的有效性,该方法充分考虑了TFN类型的数据特征和两种权重完全未知的情况,且属性信息集结更加客观高效,计算相对简便,为TFN多属性决策问题提供了新的信息集结方式和解决思路。	Finally, an example of comparative analysis to verify the effectiveness of the TFNCD operator and its assembly methods, this method fully considers the TFN characteristics and two types of data of weights is completely unknown, and attribute information gathering more objective and efficient, relatively simple calculation, TFN multiple attribute decision making problem provides a new way of information gathering and solution.