Timezone: »

MetaReg: Towards Domain Generalization using Meta-Regularization
Yogesh Balaji · Swami Sankaranarayanan · Rama Chellappa

Thu Dec 06 07:45 AM -- 09:45 AM (PST) @ Room 517 AB #157

Training models that generalize to new domains at test time is a problem of fundamental importance in machine learning. In this work, we encode this notion of domain generalization using a novel regularization function. We pose the problem of finding such a regularization function in a Learning to Learn (or) meta-learning framework. The objective of domain generalization is explicitly modeled by learning a regularizer that makes the model trained on one domain to perform well on another domain. Experimental validations on computer vision and natural language datasets indicate that our method can learn regularizers that achieve good cross-domain generalization.

Author Information

Yogesh Balaji (University of Maryland)
Swami Sankaranarayanan (Butterfly Network Inc.)
Rama Chellappa (University of Maryland College Park)

More from the Same Authors