how can i force a matrix to be positive sime-definite matrix?

Question

arkedia 2013-2-12

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/63179-how-can-i-force-a-matrix-to-be-positive-sime-definite-matrix

i'm creating a gibbs sampling algorithm and i one of thge steps an error appears in using mvnrnd function , sigma must be positive semi definite matrix what can i do?

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Jan 2013-2-12

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/63179-how-can-i-force-a-matrix-to-be-positive-sime-definite-matrix#answer_74767

编辑：Jan 2013-2-12

Please read http://en.wikipedia.org/wiki/Positive-semidefinite_matrix#Negative-definite.2C_semidefinite_and_indefinite_matrices

Perhaps it helps to know, that A*A is positive semi-definite for any matrix A. But it is not clear how your data should be changed to match the requirements. Please post more details.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

Answer 2

Youssef Khmou 2013-2-12

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/63179-how-can-i-force-a-matrix-to-be-positive-sime-definite-matrix#answer_74773

hi, arkedia,

like Jan said you can take mvnrnd(A*A) if A is square or A*conj(A') otherwise and A*A~SIGMA , but if A is an array of data then A*A will create covariance of SIGMA matrix itself which will mess your calculations,

Try to alter your code using abs(SIGMA) instead of sigma .