AI-ordlista
Den kompletta ordlistan över AI
NSGA-II
Elitist-preserving genetic algorithm using a fast non-dominated sort and a crowding distance to maintain solution diversity on the Pareto front.
ε-constraint approach
Method that transforms a multi-objective problem into single-objective optimization problems by optimizing a primary objective while constraining the others with ε thresholds.
Multi-objective trade-off
Inherent competition between conflicting objectives where improving one objective necessarily leads to the degradation of at least one other objective.
Solution archiving
Technique for storing and updating a set of non-dominated solutions throughout the optimization process to preserve the best found solutions.
Multi-objective elitism
Strategy that preserves the best solutions between generations to guarantee monotonic convergence towards the optimal Pareto front.
Indicator-based optimization
Paradigm that directly uses performance indicators, such as hypervolume, as a fitness function to guide the search towards high-quality solution sets.
Multi-objective scalability
The ability of an algorithm to maintain its performance as the number of objectives increases, often degraded by the curse of dimensionality.
Convergence and diversity
Dual criteria assessing proximity to the optimal Pareto front (convergence) and the uniform distribution of solutions on this front (diversity).
Multi-objective coevolution
Approach where multiple populations evolve simultaneously, each specialized in different regions of the Pareto front or different subsets of objectives.
Tchebychev decomposition
Weighted decomposition method transforming objectives into a scalar function using the Tchebychev norm to ensure solutions on convex and non-convex fronts.
Reference Point Approach
Interactive technique where the decision maker specifies reference points to guide the search toward specific regions of interest on the Pareto front.