A Two-Stage Bayesian Nonparametric Model for Novelty Detection with Robust Prior Information
share
Standard novelty detection methods aim at bi-partitioning the test units into already observed and previously unseen patterns. However, two significant issues arise: there may be considerable interest in identifying specific structures within the latter subset, and contamination in the known classes could completely blur the actual separation between manifest and new groups. Motivated by these problems, we propose a two-stage Bayesian nonparametric novelty detector, building upon prior information robustly extracted from a set of complete learning units. A general-purpose multivariate methodology, as well as an extension suitable for functional data objects, are devised. Some theoretical properties of the associated semi-parametric prior are investigated. Moreover, we propose a suitable ξ-sequence to construct an independent slice-efficient sampler that takes into account the difference between manifest and novelty components. We showcase our model performance through an extensive simulation study and applications on both multivariate and functional datasets, in which diverse and distinctive unknown patterns are discovered.
READ FULL TEXT