[BEP06b] Convex Quadratic Reformulation for Exact Solution of 0-1 Quadratic Programs
Conférence Internationale avec comité de lecture :
January 2006,
pp.79,
motcle:
Résumé:
We recall our general exact solution method (QCR) for linearly-constrained 0-1 quadratic programs. It uses the equality constraints in order to reformulate the objective function by an equivalent quadratic convex function. The best reformulation from the LP-relaxation point of view lies on semidefinite programming. In this communication, we present an application of this method to the densest k-subgraph problem and we give a comparison with other reformulations method. Experimental results show that, for this problem, the approach outperforms existing methods.