I have read a paper, but I can not understand one optimization thoroughly.Generally, Frobenius norm of one matrix, $\|F\|_F^2$, as the objective function is convex, so we can resolve it not using the bisection method. Within the paper, the authour say that under the given constraint, $\|F\|_F^2$ is quasiconvex, it should be resolved using the bisection method , why?
Note that, the matrix $F$ is the variable.
the paper comes from http://ieeexplore.ieee.org/xpl/articleDetails.jsp?tp=&arnumber=6555249&queryText%3DWorst-Case+Robust+Masked+Beamforming+for.
The picture is a simplified description of it.