積率行列による一般化最小2乗推定アルゴリズム

概要

論文の詳細を見る
It is well known that generalized least squares (GLS) requires a large amount of memory and time for computation, although the estimates are accurate. One way to avoid this disadvantage is a usage of modified GLS in terms of correlation function (GLSC). In the case of small data size, however, the estimated results are unreliable even for the stationary sequences postulated in GLSC.In the present paper, we propose a new GLS based on a product moment matrix to remove these defects. The proposed algorithm gives the same result as GLS but reduces computational time. It is especially effective in the case where the input is nonstationary and many iterations are required to get the good results.
公益社団法人計測自動制御学会の論文