概率型稀疏核Logistic多元分类机  

Probabilistic Sparse Kernel Logistic Multi-classifier

作　　者：郑建炜[1] 王万良[1] 蒋一波[1] 陈伟杰[1] 

机构地区：[1]浙江工业大学计算科学与技术学院,杭州310023

出　　处：《电子与信息学报》2011年第7期1632-1638,共7页Journal of Electronics & Information Technology

基　　金：国家自然科学基金(61070043);浙江省自然科学基金(Y1100611)资助课题

摘　　要：该文提出一种基于二级先验概率的多元核Logistic分类机,扩展核Logistic回归为多元模型,并解决其解的稀疏性问题,以提升多分类应用时的模型运行速率。为约简模型构建所需计算量,训练过程采用自下向上增补算法,每次迭代采用尽量少的输入样本,规避了大型矩阵逆操作,以适应于不同量度的数据场合。实验显示,所提多元分类机模型构建简单,且识别率与稀疏性都优于经典支持向量机所生成的"一对一"多分类方法及传统多元核Logistic回归算法。A new kernel logistic regression model based on two phase sparsity-promoting prior is proposed to render a sparse multi-classifier and enhance the run-time efficiency.For accelerating the building of the model,the bottom-up training algorithm is adopted which controls the capacity of the learned classifier by minimizing the number of basis functions used,resulting in better generalization and faster computation.Experimental results on standard benchmark data sets attest to the accuracy,sparsity,and efficiency of the proposed methods.

关 键 词：核Logistic回归 稀疏性概率 多元分类机 自下向上训练 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]