Towards the secrets of sota training strategies
Welcome
This repoistory is used to organize training strategies
- Regularization
- data augmentation | 2017- Luis, DL based 2018- Ekin, AutoAugment(RL Learned) 2019- Ekin, RandAugment
- dropout
- Mixup | 2017 - Zhang
- dropblock | 2018- Ghiasi, Dropblock
- improved learning rate | 2016- Ilya, SGDR; 2017- Kaiming, Large Mini-batch Learning Rate; Ilya, AdamW
- repeatitive sampling | 2019- Hoffer
- adversarial augmentation | 2020-Zhang 2020-Li