Multivariate Probability Distribution

Multivariate probability distributions model the joint probabilities of multiple variables, aiming to capture complex relationships and dependencies beyond what univariate distributions can represent. Current research emphasizes efficient representation and inference using probabilistic circuits, novel algorithms like optimistic CUCB-MT for combinatorial bandits, and generative models such as normalizing flows. These advancements improve the accuracy and efficiency of modeling multivariate data in diverse applications, ranging from reinforcement learning and electricity price forecasting to robust rotation estimation and causal inference.