YZ Sözlüğü
Yapay Zekanın tam sözlüğü
Weighted Linear Scalarization
Method combining objectives by linear weighting to reduce the multi-objective problem to a single objective.
Pareto Front
Set of non-dominated optimal solutions where no objective can be improved without degrading another one.
Constrained Reinforcement Learning
Approach that optimizes a primary objective while respecting constraints on other objectives.
Compromise Methods
Techniques that explicitly balance contradictory objectives according to defined preferences.
Multi-Objective Q-Learning
Extension of the Q-Learning algorithm handling reward vectors instead of scalar values.
Multi-Objective Evolutionary Optimization
Combination of evolutionary algorithms with RL to explore the Pareto front
Multi-Objective Hierarchical Reinforcement Learning
Hierarchical structure where different levels manage different objectives or combinations of objectives.
Multi-Objective Policies
Decision systems that produce optimal actions according to different trade-offs between objectives.
Fonctions de Valeur Multi-Objectifs
Représentations vectorielles de la valeur d'état ou d'action considérant tous les objectifs simultanément.
Deep RL Multi-Objectifs
Application de réseaux de neurones profonds pour approximer les solutions multi-objectifs complexes.
Exploration dans l'Espace des Objectifs
Stratégies d'exploration conçues pour découvrir efficacement le front de Pareto.
Apprentissage par Renforcement Multi-Agents Multi-Objectifs
Extension au multi-agents où chaque agent ou le système collectif optimise plusieurs objectifs.
Évaluation des Politiques Multi-Objectifs
Métriques et méthodes spécifiques pour évaluer et comparer les solutions multi-objectifs.
Dynamic Weight Adaptation
Methods that automatically adjust the relative importance of objectives during learning.
Continuous Multi-Objective Reinforcement Learning
Application to continuous action spaces with simultaneous optimization of multiple objectives.