How about this:
julia> cs = collect(2:N); i = 1;
julia> N=5; k=2; I = Array(Int, N*k); J = Array(Int, N*k); V = fill(true,
N*k);
julia> for r in 1:N
rng = (r-1)*k+1 : r*k
I[rng] = r
J[rng] = sample(cs, k; replace=false)
if r<N
cs[r] -= 1
end
end
julia> spmat = sparse(I, J, V, N, N)
5x5 sparse matrix with 10 Bool entries:
[2, 1] = true
[3, 1] = true
[5, 1] = true
[1, 2] = true
[4, 2] = true
[5, 2] = true
[2, 3] = true
[4, 3] = true
[3, 4] = true
[1, 5] = true
julia> full(spmat)
5x5 Array{Bool,2}:
false true false false true
true false true false false
true false false true false
false true true false false
true true false false false
On Thursday, July 23, 2015 at 2:29:36 PM UTC-4, Seth wrote:
>
> Hi all,
>
> I'm looking for really clever (FAST) ways of creating an NxN sparse random
> boolean matrix that has the following properties:
>
> * exactly k (k <= N) true values per row, and
> * no true values on the diagonal (mx[i,i]).
>
> Anyone have an ingenious way of creating something like this?
>
