Description

The course develops a coherent progression from foundational optimization principles to advanced metaheuristic techniques, beginning with the formulation of optimization problems and the limitations of classical deterministic methods, which motivates the introduction of heuristics through examples such as the Traveling Salesperson Problem and simple constructive or local-improvement procedures. Recognizing the inadequacy of heuristics for escaping local optima, the course moves to single-solution metaheuristics by first examining solution representations and neighborhood structures, then presenting gradient-based approaches, hill-climbing variants, and the role of exploration–exploitation dynamics, which sets the stage for global strategies such as Simulated Annealing, Tabu Search with memory structures and aspiration criteria, and Iterated Local Search with controlled perturbations. Population-based methods are then introduced, starting with Evolutionary Strategies and progressing to Genetic Algorithms, where encoding choices, crossover and mutation operators, selection mechanisms, elitism, and steady-state variants are examined in detail, followed by Genetic Programming and Differential Evolution, which employs vector differential operators to guide search in continuous domains. The course subsequently turns to swarm intelligence, covering Particle Swarm Optimization with velocity and neighborhood models and Ant Colony Optimization with pheromone-based probabilistic construction rules, highlighting their differences from evolutionary paradigms. It concludes with multi-objective optimization, presenting Pareto dominance, non-dominated sorting, diversity preservation, and culminating in NSGA-II, which integrates elitism, crowding mechanisms, and efficient sorting procedures to address conflicting objectives within a unified evolutionary framework.

Syllabus

The course is organized as a 30-hour program delivered through 20 sessions, each with a duration of one hour and a half. The following distribution reflects the progression of the topics presented earlier, moving from foundational concepts to advanced metaheuristic and multi-objective optimization techniques:

Session 1 — Introduction to Optimization and Problem Modeling.

Session 2 — Classical Methods and Their Practical Limitations

Session 3 — Constructive Heuristics and Local Improvement Heuristics

Session 4 — Solution Encoding and Neighborhood Structures

Session 5 — Gradient-Based Local Searcch

Session 6 — Hill Climbing and Steepest-Ascent Variants

Session 7 — Exploration–Exploitation Dynamics

Session 8 — Simulated Annealing

Session 9 — Tabu Search and Feature-Based Extensions

Session 10 — Iterated Local Search

Session 11 — Evolutionary Strategies

Session 12 — Genetic Algorithms: Encoding and Variation Operators

Session 13 — Selection Mechanisms, GA Variants, and Practical Implementation

Session 14 — Genetic Programming

Session 15 — Differential Evolution

Session 16 — Particle Swarm Optimization

Session 17 — Ant Colony Optimization

Session 18 — Pareto Optimality and Non-Dominated Sorting

Session 19 — NSGA-II

Session 20 — NSGA-II: Implementation Insights, and Course Wrap-Up

Bibliographie

Lones, Michael. "Sean Luke: essentials of metaheuristics." (2011): 333-334.

Contacts

Responsable(s) de l'enseignement

• Mahmoud Golabi

MCC

Les épreuves indiquées respectent et appliquent le règlement de votre formation, disponible dans l'onglet Documents de la description de la formation.

Régime d'évaluation

ECI (Évaluation continue intégrale)

Coefficient

2.0

Évaluation initiale / Session principale - Épreuves