An enriched mixture model for functional clustering

There is an increasingly rich literature about Bayesian nonparametric models for clustering functional observations. Most recent proposals rely on infinite-dimensional characterizations that might lead to overly complex cluster solutions. In addition, while prior knowledge the shapes typically available, its practical exploitation be a difficult modeling task. Motivated by application in e-commerce, we propose novel enriched Dirichlet mixture model data. Our proposal accommodates incorporation of constraints bounding complexity. We characterize process through urn scheme clarify underlying partition mechanism. These features very interpretable method compared available techniques. Moreover, employ variational Bayes approximation tractable posterior inference overcome computational bottlenecks.