Generalizable Label Distribution Learning

Label Distributed Learning (LDL) is an emerging machine learning paradigm that has received extensive research in recent years. Owing to the excellent capability in dealing with label ambiguity, LDL has been widely adopted in many real-world scenarios. Though remarkable progress has been achieved in various tasks, one limitation with existing LDL methods is that they are all based on the i.i.d. assumption that training and test data are identically and independently distributed. As a result, they suffer obvious performance degradation and are no longer applicable when tested in out-of-distribution scenarios, which severely limits the application of LDL in many tasks. In this paper, we identify and investigate the Generalizable Label Distribution Learning (GLDL) problem. To handle such a challenging problem, we delve into the characteristics of GLDL and find that feature-label correlation and label-label correlation are two essential subjects in GLDL. Inspired by this finding, we propose a simple yet effective model-agnostic framework named Domain-Invariant Correlation lEarning (DICE). DICE mines and utilizes the correlation between feature and label that are invariant across different domains to learn a generalizable feature-label correlation by introducing a prior alignment strategy. In the meantime, it leverages a label correlation alignment strategy to further retain the consistency of label-label correlation in different domains. Extensive experiments verify the superior performance of DICE. Our work fills the gap in benchmarks and techniques for practical GLDL problems.