search

jiequancui / ResLT

ResLT: Residual Learning for Long-tailed Recognition (TPAMI 2022)
https://arxiv.org/pdf/2101.10633.pdf
MIT License
55 stars 5 forks source link

Gamma #8

Open JingLiJJ opened 1 year ago

JingLiJJ commented 1 year ago

Thanks for your great work.

I'm confused about the gamma hyperparameter. Could you explain the function of gamma for me? Why do you just set it on the ImageNet-lt and iNaturalist datasets, not Cifar?

jiequancui commented 1 year ago

Hi,

Thanks for your interest in our work.

The gamma hyperparameter is for normalizing classifier weights, which is brought from "decoupling representation and classifier for long-tailed recognition". This technique can also be applied to CIFAR and improve performance.

Best,

JingLiJJ commented 1 year ago

Hi,

Thanks for your interest in our work.

The gamma hyperparameter is for normalizing classifier weights, which is brought from "decoupling representation and classifier for long-tailed recognition". This technique can also be applied to CIFAR and improve performance.

Best,

Thanks! I got it.