Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Deep Stable Learning for Out-Of-Distribution Generalization

About

Approaches based on deep neural networks have achieved striking performance when testing data and training data share similar distribution, but can significantly fail otherwise. Therefore, eliminating the impact of distribution shifts between training and testing data is crucial for building performance-promising deep models. Conventional methods assume either the known heterogeneity of training data (e.g. domain labels) or the approximately equal capacities of different domains. In this paper, we consider a more challenging case where neither of the above assumptions holds. We propose to address this problem by removing the dependencies between features via learning weights for training samples, which helps deep models get rid of spurious correlations and, in turn, concentrate more on the true connection between discriminative features and labels. Extensive experiments clearly demonstrate the effectiveness of our method on multiple distribution generalization benchmarks compared with state-of-the-art counterparts. Through extensive experiments on distribution generalization benchmarks including PACS, VLCS, MNIST-M, and NICO, we show the effectiveness of our method compared with state-of-the-art counterparts.

Xingxuan Zhang, Peng Cui, Renzhe Xu, Linjun Zhou, Yue He, Zheyan Shen• 2021

Related benchmarks

TaskDatasetResultRank
Domain GeneralizationVLCS
Accuracy77.65
238
Domain GeneralizationPACS
Accuracy (Art)81.74
221
Image ClassificationImageNet-A (test)--
154
Multi-class classificationVLCS
Acc (Caltech)88.25
139
object recognitionPACS (leave-one-domain-out)
Acc (Art painting)81.74
112
Image ClassificationPACS
Accuracy45.14
100
Image ClassificationVLCS
Accuracy79.15
76
Image ClassificationiWILDCam OOD
Accuracy71.52
26
Domain GeneralizationVLCS (leave-one-domain-out)
Avg Acc77.65
22
Image ClassificationDR (OOD)
Accuracy86.3
11
Showing 10 of 14 rows

Other info

Follow for update