Domain generalization (DG) methods aim to develop models that generalize to settings where the test distribution is different from the training data. In this paper, we focus on the challenging problem of multi-source zero shot DG (MDG), where labeled training data from multiple source domains is available but with no access to data from the target domain. A wide range of solutions have been proposed for this problem, including the state-of-the-art multi-domain ensembling approaches. Despite these advances, the na\"ive ERM solution of pooling all source data together and training a single classifier is surprisingly effective on standard benchmarks. In this paper, we hypothesize that, it is important to elucidate the link between pre-specified domain labels and MDG performance, in order to explain this behavior. More specifically, we consider two popular classes of MDG algorithms -- distributional robust optimization (DRO) and multi-domain ensembles, in order to demonstrate how inferring custom domain groups can lead to consistent improvements over the original domain labels that come with the dataset. To this end, we propose (i) Group-DRO++, which incorporates an explicit clustering step to identify custom domains in an existing DRO technique; and (ii) DReaME, which produces effective multi-domain ensembles through implicit domain re-labeling with a novel meta-optimization algorithm. Using empirical studies on multiple standard benchmarks, we show that our variants consistently outperform ERM by significant margins (1.5% - 9%), and produce state-of-the-art MDG performance. Our code can be found at https://github.com/kowshikthopalli/DREAME
翻译:域一般化 (DG) 方法旨在开发模型, 将测试分布与培训数据不同, 并推广到测试分布不同的环境。 在本文中, 我们集中关注多源零点拍摄 DG(MD) (MDG) (MD) (MD) (MD) (MD) (MDD) (MD) (MDD) (MD) (MDD) (MD) (MDD) (MDD) (MD) (MD) (MD) (MDG) (MD) (MDDD) (MDM) (MD) (MDR) (DR) (DR) (DR) (DR) (DR) (OD) (DR) (DR) (DR) (DR) (DR) (DR) (DR) (DR) (DR) (OD) (OD) (ODR(DR) (O) (D(DR) (D(DR) (DR(DR) (ODR) (ODR) (O) (DR) (O) (DR) (DR) (O) (O) (O) (O) (DR) (DR) (DR) (ODR) (ODR) (O) (D) (DR) (DR) (O) (O) (O) (O) (O) (DR) 。 。