YZ Sözlüğü
Yapay Zekanın tam sözlüğü
Markov Chain Monte Carlo
Class of sampling algorithms constructing a Markov chain whose stationary distribution matches the desired target distribution for probabilistic inference.
Likelihood Weighting
Weighted sampling algorithm assigning weights to samples based on the likelihood of observed variables, avoiding rejection of incompatible samples.
Laplace's Method
Gaussian approximation of a distribution around its mode, using second-order Taylor expansion of the log probability density.
Bethe Approximation
Variational method minimizing Kullback-Leibler divergence under local marginal constraints, forming the basis of belief propagation.
No-U-Turn Sampler
Advanced adaptation of Hamiltonian Monte Carlo automatically determining optimal trajectory length to optimize sampling efficiency.
Expectation Propagation
Iterative variational algorithm progressively bringing the factorized distribution closer to the target by local minimization of Kullback-Leibler divergence.