AI 용어집
인공지능 완전 사전
Continuous Differentiation Algorithm
Metaheuristic optimization technique using vector difference operations to create new candidate solutions in the hyperparameter space.
Fitness Landscape
Multidimensional representation of the objective function where each point corresponds to a hyperparameter configuration and its height to the associated performance.
Self-adaptive Mutation
Mechanism where mutation parameters evolve with solutions, allowing the algorithm to dynamically adjust its exploration intensity.
Evolutionary Niching
Technique maintaining diversity by promoting the emergence of specialized sub-populations in different regions of the hyperparameter search space.
Multi-objective Evolution
Approach optimizing multiple performance metrics simultaneously, generating a Pareto front of non-dominated hyperparameter solutions.
Recombination Operator
Mechanism combining the genetic material of multiple parents to create new individuals, exploring new hyperparameter combinations.
Hyperparameter Co-evolution
Approach where different hyperparameter sets evolve simultaneously in interaction, jointly optimizing interdependent aspects of the model.
Crossover Rate Adaptation
Mechanism dynamically adjusting the crossover probability during evolution to optimize the balance between exploitation and exploration.