2021.05.10
- Deep Domain Generalization with Feature-norm Network
主要思路几乎同”Larger Norm More Transferable: An Adaptive Feature Norm Approach for Unsupervised Domain Adaptation”，采用双网络蒸馏的方式进行泛化
- 不建议看，个人觉得有些小问题，最主要的问题：predictor有什么用？
partial domain generalization; larger feature norm; non-alignment
paper
- Invariant Risk Minimization
理论较为通用的一篇因果推断的泛化论文，强调环境之类虚假关系对预测的影响
IRM; IRMv1; casual inference; bilevel optimization; SEM; structural equation model; spurious correlation; invariant representation; optimial classifier; rank-1
paper code
- Domain Generalization Needs Stochastic Weight Averaging for Robustness on Domain Shifts
>
2021.04.17
- Semantic Segmentation with Generative Models: Semi-Supervised Learning and Strong Out-of-Domain Generalization
StyleGAN2; semi-supervised learning; SSL; out-of-distribution; Encoder
paper
2021.04.15
- Domain Generalization with MixStyle
论文本意：模拟出多种domain，采用batch特征shuffle的方式（随机打乱与域打乱，详见图2）扰动特征，然后采用实例级mixup。简单，但个人感觉不如“Open Domain Generalization with Domain-Augmented Meta-Learning”，同时shuffle直觉上有点说不过去。
特征解耦
instance mixstyle; mixup; mixstyle; feature shuffle; simple; random shuffle; domain label shuffle
paper code