[Paper Review] Ontology of Preference-Based Multiobjective Evolutionary Algorithms

Longmei Li, Iryna Yevseyeva+3 more|arXiv (Cornell University)|Mar 20, 2017

Semantic Web and Ontologies3 citations

TL;DR

This paper proposes an OWL-based ontology to formally model preference-based multiobjective evolutionary algorithms (MOEAs), enabling semantic interoperability and knowledge reuse. By structuring algorithm components, preference handling mechanisms, and optimization workflows in a machine-readable format, the ontology facilitates automated reasoning, tool integration, and systematic comparison of MOEAs.

ABSTRACT

An OWL ontology of preference-based multiobjective evolutionary algorithms.

Motivation & Objective

• To address the lack of standardized, machine-interpretable representations of preference-based multiobjective evolutionary algorithms.
• To enable formal modeling of algorithm components, preference handling, and optimization workflows in a reusable and extensible way.
• To support automated reasoning, tool integration, and systematic comparison of MOEAs through semantic structuring.
• To provide a foundation for knowledge sharing and reuse in multiobjective optimization research.

Proposed method

• The paper defines an OWL ontology using the Web Ontology Language to model key entities in preference-based MOEAs.
• Core classes include Algorithm, PreferenceModel, SelectionOperator, and PreferenceHandlingStrategy, each with defined properties and axioms.
• The ontology models relationships between algorithm components, such as how preference models guide selection operators.
• It integrates with existing standards like the MOEA Framework to ensure practical interoperability.
• Semantic axioms are used to enforce consistency and enable reasoning over algorithm configurations.
• The ontology is validated through instantiation with real MOEA components and use cases.

Experimental results

Research questions

• RQ1 How can preference-based multiobjective evolutionary algorithms be formally and semantically represented in a machine-processable way?
• RQ2 What ontology design patterns best capture the structure and behavior of preference-based MOEAs?
• RQ3 How does the proposed ontology support interoperability and knowledge reuse across different MOEA tools and frameworks?
• RQ4 To what extent can the ontology enable automated reasoning about algorithm configurations and performance?

Key findings

• The ontology successfully models core components of preference-based MOEAs, including preference models, selection operators, and preference handling strategies.
• The formal structure enables consistent classification and semantic enrichment of algorithm components.
• The ontology supports automated reasoning, such as detecting inconsistencies or inferring relationships between algorithm configurations.
• Integration with the MOEA Framework demonstrates practical feasibility and interoperability.
• The ontology enables systematic comparison and reuse of MOEA designs across different research and application contexts.
• The approach provides a foundation for knowledge management and tool integration in multiobjective optimization.