Quantifying performance inflation from hidden sample dependence in biomedical image classification benchmarks
Lukáš Jochymek, Markéta Vašinková, Petr Gajdoš
The paper quantifies how increasing train-test sample relatedness inflates classification performance in biomedical image benchmarks using a microscopy dataset.
bioRxiv · Oct 8, 2026 · PreprintBioinformatics