基于相关向量机及模拟退火的损耗模型训练  

Loss model training algorithm based on relevance vector machine combined with simulated annealing algorithm

作　　者：李丽娜[1] 马俊[1,2] 梁德骕 李文浩[1] 

机构地区：[1]辽宁大学物理学院,辽宁沈阳110036 [2]国网山东省电力公司电力科学研究院,山东济南250002

出　　处：《计算机工程与设计》2016年第1期139-145,173,共8页Computer Engineering and Design

基　　金：国家自然科学基金项目(61403176);辽宁省教育厅科学技术研究基金项目(L2013003)

摘　　要：为建立具有良好环境适应性的无线信号室内空间传播损耗模型,有效降低测距定位的误差,提出一种基于相关向量机结合模拟退火算法滤波进行室内传播损耗模型训练的算法。相关向量机选用高斯核函数来拟合样本,进行传播损耗模型训练;设计一种引入局部更新策略的模拟退火滤波算法优化相关向量机训练样本集,降低样本误差的影响。仿真结果验证了该算法的可行性及优越性,较好地满足了室内测距定位的精度要求。An indoor propagation loss model training algorithm based on relevance vector machine and simulated annealing algorithm was proposed to build the wireless signal propagation loss model of indoor space with good environment adaptability and to reduce the ranging and positioning error effectively.Gauss kernel function was used to fit the samples for the relevance vector machine propagation loss model training.And an improved simulated annealing filtering algorithm with the local updating strategy was designed to optimize the training samples of the relevance vector machine algorithm to reduce the influence of the samples error.The results of simulation show the feasibility and superiority of the proposed algorithm,which can meet the requirements of the indoor ranging and positioning accuracy well.

关 键 词：传播损耗模型 相关向量机 支持向量机 模拟退火算法 滤波 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]