AI 용어집
인공지능 완전 사전
Fisher's Z Test
Statistical test used to assess conditional independence between continuous variables under multivariate normality assumption.
Pearl's Causal Criterion
Set of formal rules enabling the identification of causal effects from graphical structures and observational data.
Markov Blanket
Minimal set of variables that, once conditioned on, render a target variable independent of all other variables in the system.
FCI Algorithm
Robust causal discovery algorithm capable of handling the presence of latent confounders and selection variables.
BIC Score
Model selection criterion penalizing structural complexity to evaluate the goodness-of-fit of a causal graph to observed data.
Equivalence Structure
CPDAG (Completed Partially Directed Acyclic Graph) representation summarizing all possible edge orientations that cannot be determined within an equivalence class.
BDeu Score
Bayesian criterion for evaluating causal graphs with discrete data, incorporating an equivalent sample prior to regularize learning.
LiNGAM Algorithm
Causal discovery method exploiting data non-Gaussianity to uniquely identify the causal ordering and graph structure.