EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
2304. An Empirical Investigation of Two Reformulation Approaches in Multiobjective Discrete Optimization
Invited abstract in session WA-37: Multiobjective Combinatorial Optimization, stream Multiobjective Optimization.
Wednesday, 8:30-10:00Room: 33 (building: 306)
Authors (first author is the speaker)
| 1. | Serpil Sayin
|
| College of Administrative Sciences and Economics, Koc University | |
| 2. | Gökhan Kof
|
| Koç University | |
| 3. | Tamey Eksi
|
| Graduate School of Sciences and Engineering, Koç University |
Abstract
We propose two alternative ways of reducing the number of objective functions of a given multiobjective discrete optimization (MODO) problem and investigate the impact of the reduction on the nondominated set of the original problem empirically. The number of objective functions is a major factor that determines the computational effort of obtaining the nondominated set of a MODO problem. Transforming the problem in a way to reduce the number of objective functions is expected to result in computational efficiencies; however, some information loss should also be exptected since the nondominated set will no longer be computed exactly. With the goal of devising methods that lead to minimal information loss, we investigate two alternative approaches: one suggests eliminating objectives, the other combining them. The elimination method uses the similarity between an objective function’s gradient and its projection onto objects defined by the remaining objectives. The combination approach relies on clustering the objective function gradients. We assess the effectiveness of the approaches on library problem instances. We report quality measures such as the coverage error, additive epsilon indicator and hypervolume to evaluate the representations obtained by solving the reduced problems. Preliminary findings indicate that proposed reductions may be effective ways to obtain quick representations of the original problem.
Keywords
- Programming, Multi-Objective
- Multi-Objective Decision Making
Status: accepted
Back to the list of papers