Learning to Balance Specificity and Invariance for In and Out of Domain Generalization

By Prithvijit Chattopadhyay, Yogesh Kumar Balaji, Judy Hoffman. 2020

239 citations recorded by the source corpus.

We introduce Domain-specific Masks for Generalization, a model for improving both in-domain and out-of-domain generalization performance. For domain generalization, the goal is to learn from a set of source domains to produce a single model that will best generalize to an unseen target domain. As such, many prior approaches focus on learning representations which persist across all source domains with the assumption that these domain agnostic representations will generalize well. However, often individual domains contain characteristics which are unique and when leveraged can significantly aid in-domain recognition performance. To produce a model which best generalizes to both seen and unseen domains, we propose learning domain specific masks. The masks are encouraged to learn a balance of domain-invariant and domain-specific features, thus enabling a model which can benefit from the predictive power of specialized features while retaining the universal applicability of domain-invariant features. We demonstrate competitive performance compared to naive baselines and state-of-the-art methods on both PACS and DomainNet.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Epidemiology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Yu Qiao
2. Ziwei Liu
3. Chen Change Loy
4. Tao A. Xiang
5. Kaiyang Zhou
6. P. Heng
7. Qi-Yan Dou
8. Jing Qin

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Chen Change Loy
3. Yu Qiao
4. Dacheng Tao
5. Xiaogang Wang
6. Ming-Hsuan Yang
7. Ziwei Liu
8. Tao A. Xiang

In Epidemiology

1. David Lopez-Paz
2. Tao A. Xiang
3. Ishaan Gulrajani
4. Qi-Yan Dou
5. Kaiyang Zhou

Citations in the corpus

• SWAD: Domain Generalization by Seeking Flat Minima (2021, neutral): “MASF [14] 82.7 - - - - - DMG [33] 83.4 - - - 43.6 - MetaReg [15] 83.6 - - - 43.6 - ER [12] 85.3 - - - - - pAdaIN [47] 85.4 - - - - - EISNet [48] 85.8 - - - - - DSON [30] 86.”
• In Search of Lost Domain Generalization (2020, background): “This augmentation protocol is increasingly standard in state-of-the-art DG work (Nam et al., 2019; Huang et al., 2020; Krueger et al., 2020; Carlucci et al., 2019a; Zhou et al., 2020; Dou et al., 2019; Hendrycks et al., 2020; Wang et al., 2020a; Seo et al., 2020; Chattopadhyay et al., 2020).”
• FedDG: Federated Domain Generalization on Medical Image Segmentation via Episodic Learning in Continuous Frequency Space (2021, background): “Domain generalization [5, 9, 12, 14, 43, 47, 58, 59] aims to learn a model from multiple source domains such that it can directly generalize to unseen target domains.”
• Learning to Diversify for Single Domain Generalization (2021, background): “FCR [25], MASF [10] and DMG [6], requires domain identifications in the training phase.”
• Domain Generalization: A Survey (2021, methodology): “Minimizing the KL Divergence As a commonly used distribution divergence measure, the KL divergence has also been employed for domain alignment [172], [173].”