Analyzing Dominance Move (MIP-DoM) Indicator for Multi- and Many-objective Optimization

TL;DR

This paper introduces a mixed integer programming approach to efficiently compute the dominance move (DoM) indicator for multi- and many-objective optimization, overcoming previous computational limitations especially in higher dimensions.

Contribution

The work proposes a novel MIP-based method to calculate DoM for three or more objectives, extending its applicability beyond bi-objective cases and addressing computational challenges.

Findings

MIP-DoM accurately computes DoM in bi-objective cases.

The method scales to higher dimensions, handling up to 30 objectives.

Experiments demonstrate the model's correctness and efficiency.

Abstract

Dominance move (DoM) is a binary quality indicator that can be used in multi-objective and many-objective optimization to compare two solution sets obtained from different algorithms. The DoM indicator can differentiate the sets for certain important features, such as convergence, spread, uniformity, and cardinality. DoM does not use any reference, and it has an intuitive and physical meaning, similar to the $\epsilon$-indicator, and calculates the minimum total move of members of one set so that all elements in another set are to be dominated or identical to at least one member of the first set. Despite the aforementioned properties, DoM is hard to calculate, particularly in higher dimensions. There is an efficient and exact method to calculate it in a bi-objective case only. This work proposes a novel approach to calculate DoM using a mixed integer programming (MIP) approach, which can handle sets with three or more objectives and is shown to overcome the $\epsilon$-indicator's information loss. Experiments, in the bi-objective space, are done to verify the model's correctness. Furthermore, other experiments, using 3, 5, 10, 15, 20, 25 and 30-objective problems are performed to show how the model behaves in higher-dimensional cases. Algorithms, such as IBEA, MOEA/D, NSGA-III, NSGA-II, and SPEA2 are used to generate the solution sets (however any other algorithms can also be used with the proposed MIP-DoM indicator). Further extensions are discussed to handle certain idiosyncrasies with some solution sets and also to improve the quality indicator and its use for other situations.