PPL
Enrique S. Quintana-Ortí
Papers (5)
StableGrad: Backward Scale Control without Batch NormalizationFedOUI: OUI-Guided Client Weighting for Federated AggregationOUI as a Structural Observable: Towards an Activation-Centric View of Neural Network TrainingOUIDecay: Adaptive Layer-wise Weight Decay for CNNs Using Online Activation PatternsRefresh-Scaling the Memory of Balanced Adam