Tutorials

Tutorial (IEEE WCCI 2026)

Introduction to Multi-Objective Optimization
IEEE WCCI 2026 (IJCNN)
21-26 June, 2026, Maastricht, the Netherlands

Short introduction:
In numerous real-world scenarios, problem-solving tasks often involve multiple conflicting objectives. For example, in neural architecture search (NAS), the goal is to automatically find the best architecture of a neural network that balances multiple objectives, such as maximizing accuracy while minimizing the model complexity. This is inherently a multi-objective problem because improving one objective (e.g., accuracy) often leads to a deterioration of another (e.g., model complexity). Traditional methods, such as the weighted sum approach, are often used but have significant drawbacks: the choice of weights is subjective, and multiple runs are required to find a set of trade-off solutions. This tutorial will guide attendees from the fundamentals of multi-objective optimization to modern, population-based solutions. First, we will use practical examples to clearly demonstrate the differences between single-objective and multi-objective optimization. Next, we explain some basic concepts of multi-objective optimization such as Pareto dominance relation, Pareto optimal solutions, Pareto fronts, and non-dominated solutions. Then, we explain traditional approaches in the MCDM (multi-criteria decision making) field where a single final solution is found. Finally, we explain population-based approaches in the EMO (evolutionary multi-objective optimization) field. We will also briefly explain some hot research topics in the EMO field.

Outline of the tutorial:
The duration of the tutorial will be two hours. It will be composed of the following parts:
Part 1. Introduction
• Real-world problems with multiple conflicting objectives
  — Examples: Neural architecture search, resource allocation
• From “finding the best” to “finding trade-offs”
• Structure and goals of this tutorial

Part 2. Fundamentals of Multi-Objective Optimization
• Basic definitions and mathematical formulation
• Pareto dominance relation and Pareto optimality
• Non-dominated solutions and the Pareto front
• Visualization examples

Part 3. Traditional Multi-Criteria Decision Making (MCDM) Approaches
• Weighted-sum method
• ε-constraint and goal programming approaches
• Strengths and limitations (subjective weights, non-convexity issues)

Part 4. Evolutionary Multi-Objective Optimization (EMO)
• Population-based search and the concept of Pareto set approximation
• General EMO workflow
  — Initialization
  — Variation (crossover/mutation)
  — Selection and environmental selection
  — Archiving and elitism
• Examples of representative algorithms
  — NSGA-II, MOEA/D, SMS-EMOA, NSGA-III

Part 5. Emerging Research Topics and Open Challenges
• Many-objective optimization
• Preference-based EMO
• Indicator-based selection and performance assessment (HV, IGD, etc.)
• Applications in machine learning and neural architecture search

Part 6. Summary and Q&A

Speakers:
Lie Meng Pang, Southern University of Science and Technology, China.
Lie Meng Pang received her Bachelor of Engineering degree in Electronic and Telecommunication Engineering and Ph.D. degree in Electronic Engineering from the Faculty of Engineering, Universiti Malaysia Sarawak, Malaysia, in 2012 and 2018, respectively. She is currently a research associate with the Department of Computer Science and Engineering, Southern University of Science and Technology (SUSTech), China. Her current research interests include evolutionary multi-objective optimization and fuzzy systems. She is a Senior Member of the IEEE and a member of the IEEE Computational Intelligence Society (CIS) Evolutionary Computation Technical Committee. She received the Best Paper Award at EMO 2025.