The title is self-evident, I think. For example, we could generate a $n \times n$ random rank-$1$ positive semidefinite matrix by generating a random vector $x$ and $x x^T$ will be the random matrix we want. Now if there is a given linear matrix inequality, i.e., a spectrahedron, is there a way to generate a matrix that is included in the spectrahedron?

So to make it more clear. Consider symmetric matrix space $S^n$,we can generate a random positive definite matrix $X$ in this space by, for example, generate some matrix $A$ and let $X=A^TA$,but what if I want to generate a random positive definite matrix in a certain subspace of the matrix space, such as defined by a Linear Matrix Inequality (as Omnomnomnom suggested in the comments). Is there a way to do it?

Thank you!

  • 1
    $\begingroup$ This could be made clearer. $\endgroup$
    – parsiad
    Commented Jul 5, 2015 at 16:05
  • $\begingroup$ I suppose you're specifically referring to linear matrix inequalities $\endgroup$ Commented Jul 5, 2015 at 16:06

1 Answer 1


Haven't dig into details but judging from the title it should be what I am looking for. Or at least give me a understanding of what can be achieved.Uniform sampling in semi-algebraic sets

  • $\begingroup$ Put your answer (that is not an answer) into your above question. $\endgroup$
    – user91684
    Commented Jun 20, 2016 at 10:26

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .