Combining the interior-point algorithm for linear programming,
which partly makes use of fast matrix multiplication, and the
generalized nested dissection technique, it is shown that the
planar minimum cost flow problem can be solved more efficiently
than existing algorithms.  This is quite interesting, since a
nonlinear approach really gives a better result for combinatorial
problems in some case.