arXiv:2402.16388v4 Announce Type: replace 
Abstract: The need for uncertainty quantification in anomaly detection systems has become increasingly important. In this context, effectively controlling Type I error rates without inflating Type II error rates in these systems can build trust and reduce costs associated with false discoveries. The field of conformal anomaly detection emerges as a promising approach for providing respective statistical and finite-sample validity guarantees through model calibration. However, reliance on calibration data imposes practical limitations, especially in low-data regimes. In this work, we formally define and evaluate leave-one-out-, bootstrap-, and cross-conformal methods for conformal anomaly detection, building on methods from the field of conformal prediction. Looking beyond the classical split-conformal approach, we show that derived methods for calculating resampling-conformal $p$-values offer a practical compromise between the data efficiency of full-conformal (transductive) approaches and the computational efficiency of split-conformal (inductive) methods. We validate derived methods and quantify their improvements for a range of one-class classifiers and datasets.

قدمت دراسة حديثة طرقًا جديدة للكشف عن الشذوذ تعتمد على أساليب leave-one-out وbootstrap وcross-conformal، مما يبرز أهمية قياس عدم اليقين في هذه الأنظمة. تؤكد الأبحاث على ضرورة التحكم في معدلات الخطأ من النوع الأول دون زيادة معدلات الخطأ من النوع الثاني، وهو أمر حاسم لبناء الثقة في تطبيقات الكشف عن الشذوذ.

Un estudio reciente ha introducido métodos de detección de anomalías basados en leave-one-out, bootstrap y conformidad cruzada, enfatizando la importancia de la cuantificación de la incertidumbre en estos sistemas. La investigación destaca la necesidad de controlar las tasas de error de tipo I sin inflar las tasas de error de tipo II, lo cual es crucial para generar confianza en las aplicaciones de detección de anomalías.

Une étude récente a introduit des méthodes de détection d'anomalies basées sur le leave-one-out, le bootstrap et la conformité croisée, soulignant l'importance de la quantification de l'incertitude dans ces systèmes. La recherche met en avant la nécessité de contrôler les taux d'erreur de type I sans gonfler les taux d'erreur de type II, ce qui est crucial pour instaurer la confiance dans les applications de détection d'anomalies.

A recent study has introduced leave-one-out, bootstrap, and cross-conformal methods for anomaly detection, emphasizing the importance of uncertainty quantification in these systems. The research highlights the need to control Type I error rates without inflating Type II error rates, which is crucial for building trust in anomaly detection applications.