超参数设置 - Githubissues

Alibaba-MIIL / ASL

Official Pytorch Implementation of: "Asymmetric Loss For Multi-Label Classification"(ICCV, 2021) paper

MIT License

732 stars 102 forks source link

Closed JNUpython closed 2 years ago

JNUpython commented 2 years ago

针对mutil label的数据不均衡程度，对ASL 的loss超参数有没有一个推荐配置了

mrT23 commented 2 years ago

this question is too general.

as a rule of thumb, the large the imbalancing, you need larger asymmetry in ASL.