Regularization for Fine-tuning
A regularization which can significantly enhance the performance of any fine-tuning methods when there is a limited amount of downstream data available. It can be used on multiple networks (ResNet, VGG, MobileNet and ViT) and fine-tuning methods (Full fine-tuning, Linear probing, L2-SP, DELTA, Surgical tuning and LoRA).
Hao ZHENG
Ph.D. candidate of AI
My research interests include transfer learning, few-shot classification and representation learning.