k-Clustering Minimum Biclique Completion via a Hybrid CP and SDP Approach

Authors:
Stefano Gualandi
Affiliations:
Dipartimento di Elettronica e Informazione, Politecnico di Milano,
Venue:
CPAIOR '09 Proceedings of the 6th International Conference on Integration of AI and OR Techniques in Constraint Programming for Combinatorial Optimization Problems
Year:
2009

Citing 7
Cited 0

Improved approximation algorithms for maximum cut and satisfiability problems using semidefinite programming

Journal of the ACM (JACM)
Cost-Based Domain Filtering

CP '99 Proceedings of the 5th International Conference on Principles and Practice of Constraint Programming
Simple and Fast: Improving a Branch-And-Bound Algorithm for Maximum Clique

ESA '02 Proceedings of the 10th Annual European Symposium on Algorithms
Exploiting semidefinite relaxations in constraint programming

Computers and Operations Research
Algorithm 875: DSDP5—software for semidefinite programming

ACM Transactions on Mathematical Software (TOMS)
The power of semidefinite programming relaxations for MAX-SAT

CPAIOR'06 Proceedings of the Third international conference on Integration of AI and OR Techniques in Constraint Programming for Combinatorial Optimization Problems
Biclique completion problems for multicast network design

Discrete Optimization

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a hybrid Constraint Programming (CP) and Semidefinite Programming (SDP) approach to the k -clustering minimum biclique completion problem on bipartite graphs. The problem consists in partitioning a bipartite undirected graph into k clusters such that the sum of the edges that complete each cluster into a biclique, i.e. , a complete bipartite subgraph, is minimum. The problem arises in telecommunications, in particular in bundling channels in multicast transmissions. In literature, the problem has been tackled with an Integer Bilinear Programming approach. We introduce two quasi-biclique constraints and we propose a SDP relaxation of the problem that provides much stronger lower bounds than the Bilinear Programming relaxation. The quasi-biclique constraints and the SDP relaxation are integrated into a hybrid CP and SDP approach. Computational results on a set of random instances provide further evidence about the potential of CP and SDP hybridizations.