March 7, 2026Dropout as Shared-Parameter BaggingA concrete way to understand dropout in deep learning: sampled subnetworks with shared weights, standing in for a much more expensive ensemble.
