Comparison and Bayesian Estimation of Feature Allocations

Feature allocation models postulate a sampling distribution whose parameters are derived from shared features. Bayesian place prior on the feature allocation, and Markov chain Monte Carlo is typically used for model fitting, which results in thousands of allocations sampled posterior distribution. Based these samples, we propose method to provide point estimate latent allocation. First, introduce FARO loss, function between satisfies quasi-metric properties allows comparing with differing numbers The loss involves finding optimal ordering among all possible orderings, but computational feasibility achieved by framing this task as linear assignment problem. We also FANGS algorithm obtain Bayes minimizing expected using available samples. can produce an other than those visited chain. investigation existing methods our proposed methods. Our search implemented fangs package R.