Problem with cvx/sparse function

I tried to use the cvx/sparse function to create a indicator matrix from the variable vector x (see below (full( sparse(1:n,x,1,n,k))).

cvx_begin
cvx_solver Mosek
variable x(n) integer
minimize( sum_square(freshape(full( sparse(1:n,x,1,n,k))’,nm,1)-B) )
subject to
A*x<=b
LB <= x <= UB
cvx_end

But I got the following error:

Error using cvx/sparse (line 25)
The first two arguments must be numeric.

The second argument in the cvx/sparse function is the cvx matrix “x”. Its elements are numeric. How can I solve this problem?

Besides, the third argument in the cvx/sparse function is a vector (in my example above, it’s a vector of all 1’s) rather than a structure. So I got the following error. How can I solve it?

Attempt to reference field of non-structure array.

Error in cvx/sparse (line 46)
[ ix, jx, vx ] = find( x.basis_ );

Thank you very much!

http://ask.cvxr.com/t/why-isnt-cvx-accepting-my-model-read-this-first/570/2?source_topic_id=2801