Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

EfficientNet解説 #346

Open
AkihikoWatanabe opened this issue May 24, 2021 · 1 comment
Open

EfficientNet解説 #346

AkihikoWatanabe opened this issue May 24, 2021 · 1 comment

Comments

@AkihikoWatanabe
Copy link
Owner

https://qiita.com/omiita/items/83643f78baabfa210ab1

@AkihikoWatanabe
Copy link
Owner Author

AkihikoWatanabe commented May 24, 2021

既存画像認識モデルの構造は変化させず、広さ、深さ、解像度を複合スケーリングすることで、従来よりも少ないパラメータ数、かつ学習速度でSoTAを達成。広さ、深さ、解像度はそれぞれ性能に互いに影響しあっており、従来のように別々にスケーリングするのではなく、3つのバランスをとりながらスケーリングする。スケーリングする際は、結果的にはそれぞれをある値で定数倍すれば良く、そのある値は最大メモリや最大FLOPS数以下(およびFLOPSが2のΦ乗で増加するような)といった制約下でAccuracyが最大化される値をグリッドサーチで見つける(らしい。ざっくりとした理解)。
転移学習しても多くのタスクでSoTA達成した。

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

1 participant