Recurrent connections facilitate occluded object recognition by explaining-away
Byungwoo Kang, Benjamin Midler, Feng Chen, Shaul Druckmann
This paper shows recurrent neural networks can improve occluded object recognition by using occluder information to 'explain-away' occlusions, as demonstrated through computational models and human experiments.
Nature Communications · Jan 31, 2026Peer-reviewedNeuroscience