Glossario IA
Il dizionario completo dell'Intelligenza Artificiale
Variational MCMC
Hybridization between Markov Chain Monte Carlo methods and variational inference to accelerate convergence while preserving exact sampling from the posterior distribution.
Stochastic Gradient MCMC
Combination of stochastic gradient methods with MCMC to perform large-scale inference on massive datasets while guaranteeing valid sampling properties.
Integrated Autocorrelation Time
Measure of MCMC algorithm efficiency quantifying the number of steps required to obtain approximately independent samples from the chain.
Burn-in Period
Initial sampling phase of MCMC during which iterations are discarded to allow the chain to converge to its stationary distribution.
Adaptive MCMC
MCMC methods dynamically adjusting their proposal parameters based on the chain history to optimize acceptance rates and accelerate convergence.
Posterior Predictive Check
Bayesian model validation comparing observed data with replicated data generated from the posterior predictive distribution.
Effective Sample Size (ESS)
Number of equivalent independent samples provided by a correlated MCMC chain, essential metric for evaluating inference quality.