摘要
在微分几何框架下,将平稳时间序列数据嵌入到曲指数流形中,从几何角度建立参数的极大似然估计与流形曲率的联系,研究MLE的渐近性态与信息损失问题,并讨论了信息的恢复和减少这一损失的参数变换方法。
Based on the differential geometry, stationary time series data is modeled as the curved exponential family. In the view of geometry. the MLE of the unknown parameters is analyzed associated with embedded curvatures and information loss in this summary process is investigated. Meanwhile a information recovery method and parameter -transformation are discussed as well.
出处
《控制与决策》
EI
CSCD
北大核心
1997年第4期307-311,共5页
Control and Decision
基金
国家自然科学基金资助课题
关键词
曲率
信息损失
ARMA数据
数理统计
curvature, information loss, observed information, ARMA model