I am doing something like the following but that runs the optimizer several times, which is slow.
for i = 1:nDimY cvx_begin quiet variables W_y(nDimY) W_x(nDimX) minimize (norm(W_y'*trout - W_x'*trin,2)) subject to W_y(i) == 1 cvx_end end
Result is the minimum of the loop. Assuming that 0 <= W_y(i) <=1 then I can write
subject to norm(W_y,Inf) == 1
as a convex optimization task
but that constraint is not allowed. Can I reword this constraint in a way CVX will appreciate?