ID 原文 译文
58168 提出一种毫米波大规模多输入多输出( MIMO) 系统中基于几何均值分解( GMD) 的混合预编码方案. In millimeter wave massive multiple input multiple output ( MIMO) systems，a hybrid precoding scheme based on geometric mean decomposition ( GMD) was presented.
58169 通过GMD 处理将信道分解为等增益的子信道，以简化编解码复杂度. In the proposed scheme，thechannel was decomposed into multiple equal-gain subchannels by means of GMD to simplify the complexity of encoding and decoding.
58170 在此基础上，推导出基于 GMD 的系统频效优化目标函数解析式; Based on it，the analytical expression of the objective function of systemspectrum efficiencyoptimization was derived.
58171 然后根据基追踪原理和最小二乘法分别设计模拟预编码和数字预编码; And then the hybrid precoding was designed according tothe basic tracking principle and the least square method.
58172 最后通过相应的优化算法得到系统频效的优化解. Finally，the optimal theoretical value of the system spectrum efficiency was obtained by the proposed algorithm.
58173 数值仿真结果表明，提出的基于 GMD 的混合预编码方法与正交匹配追踪的方案相比，能明显降低系统编解码复杂度，并提高系统频效. The numerical simulation results showthat the proposed scheme has the advantages on reducing system complexity and improving system spectrum efficiency compared with the design scheme based on orthogonal matching pursuit.
58174 针对移动边缘计算( MEC) ，提出了一种基于机器学习的随机任务迁移算法，通过将任务划分为可迁移组件和不可迁移组件，结合改进的 Q 学习和深度学习算法生成随机任务最优迁移策略，以最小化移动设备能耗与时延的加权和. For mobile-edge computing ( MEC) ，a machine learning-based stochastic task offloading algorithm was proposed. By dividing the task into offloadable components and unoffloadable components，theimproved Q learning and deep learning algorithm were used to generate the optimal offloading strategy ofstochastic task，which minimized the weighted sum of energy consumption and time delay of the mobiledevices.
58175 仿真结果表明，该算法的时延与能耗加权和与移动设备本地执行算法相比节约了 38. The simulation results show that the proposed algorithm saves the weighted sum of energy consumption and time delay by 38.
58176 1% . 1% ，compared to the local execution algorithm.
58177 基于聚类算法的混合分类器构建的信息评分系统中，不合理的聚类值或者初始类簇中心点会严重影响分类精度的问题，对此，提出了 2 种基于模糊粗糙集实例选择的新型混合算法. For the credit scoring system built on cluster algorithm based hybrid classifier，the unreasonable clusters number or starting center points of each cluster have severely negative influence on the classification accuracy.