相加的血縁行列の逆行列を用いた傾斜勾配法による混合モデル方程式の解法の比較

書誌事項

タイトル別名
  • Comparison of Algorithms with Setting up Inverse of Additive Relationship Matrix for Solving Mixed Model Equations Using Scaled Conjugate Gradient Iteration.
  • ソウカテキ ケツエン ギョウレツ ノ ギャクギョウレツ オ モチイタ ケイシャ コウバイホウ ニ ヨル コンゴウ モデル ホウテイシキ ノ カイホウ ノ ヒカク

この論文をさがす

抄録

A semi-indirect approach, where not the whole equations but inverse of additive relationship matrix was set up, was proposed in order to compute the solutions to mixed model equations. The semi-indirect and direct algorithms for solving mixed model equations using scaled conjugate gradient iteration were compared for 8 data sets generated by the Monte Carlo simulation. Data sets 1 through 4 contained a single dependent variable with about 2, 000, 000 animals. Data sets 5 through 8 consisted of five traits with about 400, 000 animals. The ordering of mixed model equations was fixed effects followed by additive genetic random effects. In direct algorithm implemented with single precision, the scaled conjugate gradient program did not converge for several data sets. The CPU time using the new technique was not different from that by the direct approach. The semi-indirect approach required smaller size of memory than the direct approach. Advantage of semiindirect approach depends on the model; more complicated models will produce clearer differences.

収録刊行物

被引用文献 (1)*注記

もっと見る

参考文献 (13)*注記

もっと見る

詳細情報 詳細情報について

問題の指摘

ページトップへ