From Imbalance to Evidence: Deep Nonparametric Testing for Complex Data under Severe Sample Imbalance

The paper introduces EIAD-C2ST, a deep classifier two-sample test designed for highly imbalanced data, particularly complex non-Euclidean structures like graphs. It reframes sample imbalance as a source of repeated testing evidence rather than a nuisance, tailoring classifiers to data structure and offering a scalable representation‑amortized strategy for expensive deep models. The authors provide theoretical guarantees on asymptotic size and power consistency, and demonstrate empirical improvements through simulations and a blockchain forensics application.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.