There are a couple of formulations in Minimize log(1+1/x) where 0<x<inf . Use `z/b`

in place of `x`

, and divide everything by `log(2)`

to make it `log2`

instead of `log`

. You can get the same argmax without dividing by `log(2)`

.

The problem as stated will be considered as ill-posed, because, assuming b > 0, the optimal solution is z = \infty, which the solvers CVX call will never get to.