CoMISo - Constrained Mixed-Integer Solver. This is the project page of the Constrained Mixed-Integer Solver (short CoMISo) developed at the Computer Graphics Group of the RWTH Aachen University, which emerged from our Mixed-Integer Quadrangulation project. This software provides a convenient tool for setting up and solving linear systems stemming from quadratic energies subject to linear equality constraints as well as to integer constraints. Basically only the system matrix, the constraints and the variables to be rounded are provided to the solver which returns a properly indexed solution vector. This drastically minimizes the programmer’s effort as a proper elimination of the constraints and the necessary reindexing is taken care of internally. The difference to popular constraining techniques such as Lagrangian Multipliers, which require no reindexning (but unfortunately result in larger, non-positive semidefinite systems), is that CoMISo efficiently eliminates the constraints from the system matrix, yielding a smaller system while at the same time preserving positive semidefiniteness.

Keywords for this software

Anything in here will be replaced on browsers that support the canvas element