Recently, a number of graph partitioning applications have emerged with additional requirements that the traditional graph partitioning model alone cannot effectively handle. One such class of problems is those in which multiple objectives, each of which can be modeled as a sum of weights of the edges of a graph, must be simultaneously optimized. This class of problems can be solved utilizing a multi-objective graph partitioning algorithm. We present a new formulation of the multi-objective graph partitioning problem and describe an algorithm that computes partitionings with respect to this formulation. We explain how this algorithm provides the user with a fine-tuned control of the tradeoffs among the objectives, results in predictable partitionings, and is able to handle both similar and dissimilar objectives. We show that this algorithm is better able to find a good tradeoff among the objectives than partitioning with respect to a single objective only. Finally, we show that by modifying the input preference vector, the multi-objective graph partitioning algorithm is able to gracefully tradeoff decreases in one objective for increases in the others.
[1]
Peter C. Fishburn,et al.
Decision And Value Theory
,
1965
.
[2]
John B. Kidd,et al.
Decisions with Multiple Objectives—Preferences and Value Tradeoffs
,
1977
.
[3]
Marek Makowski,et al.
Methodology and a Modular Tool for Multiple Criteria Analysis of LP Models
,
1994
.
[4]
Martin G. Everett,et al.
Parallel Dynamic Graph Partitioning for Adaptive Unstructured Meshes
,
1997,
J. Parallel Distributed Comput..
[5]
Vipin Kumar,et al.
A Parallel Algorithm for Multilevel Graph Partitioning and Sparse Matrix Ordering
,
1998,
J. Parallel Distributed Comput..
[6]
Vipin Kumar,et al.
Multilevel Algorithms for Multi-Constraint Graph Partitioning
,
1998,
Proceedings of the IEEE/ACM SC98 Conference.