We introduce a powerful and non-random methodology for novelty detection, offering distribution-free false discovery rate (FDR) control guarantees. Building on the full-conformal inference framework and the concept of e-values, we introduce full-conformal e-values to quantify evidence for novelty relative to a given reference dataset. These e-values are then utilized by carefully crafted multiple testing procedures to identify a set of novel units out-of-sample with provable finite-sample FDR control. Furthermore, our method is extended to address distribution shift, accommodating scenarios where novelty detection must be performed on data drawn from a shifted distribution relative to the reference dataset. In all settings, our method is non-random and can perform powerfully with limited amounts of reference data. Empirical evaluations on synthetic and real-world datasets demonstrate that our approach significantly outperforms existing methods for novelty detection.
翻译:暂无翻译