🤖 AI Summary
In few-shot anomaly detection, simultaneously controlling Type-I error rate (α) and achieving high statistical power (1−β) remains challenging. To address this, we propose three resampling-based conformal p-value methods—leave-one-out, bootstrap, and cross-conformal—which strictly control α without requiring an independent calibration set. We formally define and systematically evaluate this class of conformal anomaly detectors for the first time, thereby overcoming the conventional inductive conformal prediction’s reliance on additional calibration data. Our framework is compatible with standard one-class classifiers (e.g., OC-SVM, Isolation Forest), ensuring both statistical rigor and computational efficiency. Experiments across multiple benchmark datasets demonstrate that our approach improves statistical power by 12–28% over split-conformal methods, while drastically reducing computational overhead compared to full-conformal inference. This enables statistically guaranteed and practically viable anomaly detection in low-data regimes.
📝 Abstract
The requirement of uncertainty quantification for anomaly detection systems has become increasingly important. In this context, effectively controlling Type I error rates (α) without compromising the statistical power (1−β) of these systems can build trust and reduce costs related to false discoveries. The field of conformal anomaly detection emerges as a promising approach for providing respective statistical guarantees by model calibration. However, the dependency on calibration data poses practical limitations — especially within low-data regimes. In this work, we formally define and evaluate leave-one-out-, bootstrap-, and cross-conformal methods for anomaly detection, incrementing on methods from the field of conformal prediction. Looking beyond the classical inductive conformal anomaly detection, we demonstrate that derived methods for calculating resampling-conformal p-values strike a practical compromise between statistical efficiency (full-conformal) and computational efficiency (split-conformal) as they make more efficient use of available data. We validate derived methods and quantify their improvements for a range of one-class classifiers and datasets.