SaluNet: no normalization layers needed with learnable activation

Arxiv CS.CV (computer vision)

Paper shows normalization layers can be fully replaced by a single learnable activation mechanism. The method, SaluNet, achieves total plasticity in deep networks without BatchNorm or LayerNorm.

deep-learning

Arxiv CS.CV (computer vision)·Mourad Zaied (University of GabesTuisia)

8 days ago