Abstract:
We consider the class of all $n\times n$$(0,1)$-matrices with
$r$ ones in each row, $2\le r\le n$.
For a matrix $P$ chosen randomly and equiprobably from this class, we present
sufficient conditions under which the law of large numbers
for the permanent $\operatorname{per}P$ is valid in the triangular array scheme
as $n\to\infty$ and the parameter
$r=r(n)\to\infty$ so that $\sqrt{n}=o(r)$.
A similar problem is solved for random $n\times n$ stochastic matrices
whose rows are independent $n$-dimensional random variables
which are identically distributed by the Dirichlet law
with parameter $\nu$ under the condition that
$n\to\infty$ and the parameter $\nu=\nu(n)>0$ varies so that $n\nu^2\to\infty$.