next up previous contents
Next: Lovász function Up: Specialized Routines Previous: Specialized Routines

Diagonally constrained problems

  For the case of diagonally constrained problems (for example, MAX--CUT relaxations), the Schur complement equations can be formed and solved very efficiently using the XZ search direction [2,3] (sometimes called the KSH/HRVW/M direction in the literature.) The specialized routines dsdp.m and fdsdp.m take advantage of this. As before, the five steps involved in setting up and solving a problem are: preparing the data, setting the parameters, initializing the variables, invoking the solver, and interpreting the output. These are almost identical to the description in Section 3, except for the following important differences:



next up previous contents
Next: Lovász function Up: Specialized Routines Previous: Specialized Routines



Madhu Nayakkankuppam
Fri Mar 28 00:48:56 EST 1997