search

yaoyao-liu / meta-transfer-learning

TensorFlow and PyTorch implementation of "Meta-Transfer Learning for Few-Shot Learning" (CVPR2019)
https://lyy.mpi-inf.mpg.de/mtl/
MIT License
731 stars 147 forks source link

About SS operation. #60

Open biwanqing opened 1 year ago

biwanqing commented 1 year ago

Hello, thanks for your very enlightening work. Could you please explain the rationale for fine-tuning the SS factor instead of the weights and bias?Are there any experimental phenomena or theoretical derivations that support this?

yaoyao-liu commented 1 year ago

Thanks for your interest in our work!

The experiments in Table 1 of our paper show updating SS performs better in few-shot learning. Our paper is based on empirical results, and there are no theoretical derivations.

If you have any further questions, please do not hesitate to contact me.

biwanqing commented 1 year ago

Thanks for your prompt reply. I am interested in this work and will continue to follow it.