Would it be possible to generate samples $x \in \mathbb{R}^n$ from $\mathcal{N}(\mu, \Sigma)$ subject to an inequality constraint $x^\top Q x/2 + b^\top x \le c$, $Q = Q^\top \succeq 0$. We also know, that and $\mu^\top Q \mu/2 + b^\top \mu < c$. Thank you!
Asked
Active
Viewed 31 times
3
-
what about just using rejection sampling? – jld Sep 19 '19 at 16:36
-
1@jld Yes, that's natural. The problem is that for $n\ge 2$ it could be grossly inefficient unless executed very cleverly. The problem worsens as $n$ increases. – whuber Sep 19 '19 at 16:56