[comp.theory] a problem in random graph

olivier@ut-emx.uucp (Olivier Goldschmidt) (05/02/91)

I'd like to know if you have some information on the following problem.
Let me first define a probability space for directed random graphs.  A graph
drawn from the probability space G(k-in,k-out) is a labelled directed
graph on n vertices.  For every vertex i, we direct an arc from i to k vertices
selected at random among the (n-1) other vertices and direct an arc from k 
vertices selected at random to i.  
Note that an arc (i,j) can be selected twice, once as an out-arc of
i and once as an in-arc of j.  The in-degree as well as the out-degree of a
vertex is at least k, and the graph contains at most 2kn arcs.
I am interested in a lower bound on k, say b(n), such that if k>=b(n), then
a graph drawn from G(k-in,k-out) posesses a Hamiltonian cycle with high
probability.  Do you know of any such result?  Thanks for your help.
Olivier.