YZ Sözlüğü
Yapay Zekanın tam sözlüğü
Random node
Random variable represented by a vertex in a Bayesian network, characterized by its state space and its conditional probabilities given its parents.
Directed arc
Oriented arrow in a Bayesian network indicating a direct dependency relationship from a parent node to a child node, representing a causal or statistical influence.
D-separation
Graphical criterion for determining conditional independence between sets of nodes in a Bayesian network by analyzing undirected paths and their blocking by observed nodes.
EM algorithm for Bayesian networks
Method for estimating parameters of Bayesian networks with incomplete data, alternating between an expectation (E) phase and a maximization (M) phase to maximize likelihood.
Partition function
Normalization constant in the joint probability distribution of a Bayesian network, ensuring that the sum of probabilities equals 1 over all possible configurations.
Noisy-OR
Compact parametric model for binary conditional distributions with multiple causes, exponentially reducing the number of parameters required in CPTs.
Marginal independence
Property where two variables are independent in their joint distribution without conditioning on other variables, which can be detected by D-separation in the moralized graph.
Variable elimination inference
Exact inference technique that sequentially eliminates variables from the network by aggregating factors, with complexity dependent on the treewidth of the graph.