SURGE: Surrogate Gradient Adaptation in Binary Neural Networks

SURGE proposes learnable gradient compensation for binary neural network training to address gradient mismatch and information loss in straight-through estimators and fixed-range clipping methods.