目标条件表示独立性(TCRI):从领域不变到领域通用表示
Target Conditioned Representation Independence (TCRI); From Domain-Invariant to Domain-General Representations
摘要:
我们提出了一个目标条件表示独立性(TCRI)目标的领域泛化。TCRI解决了现有的领域泛化方法由于不完整的约束的局限性。具体来说,TCRI实现正则化动机的条件独立约束,足以严格学习完整的不变机制,我们证明是必要的和充分的领域泛化。经验上,我们表明,TCRI是有效的合成和真实世界的数据。TCRI在平均精度方面与基线具有竞争力,而在最差域精度方面优于基线,这表明所需的跨域稳定性。
英文摘要:
We propose a Target Conditioned Representation Independence (TCRI) objective for domain generalization. TCRI addresses the limitations of existing domain generalization methods due to incomplete constraints. Specifically, TCRI implements regularizers motivated by conditional independence constraints that are sufficient to strictly learn complete sets of invariant mechanisms, which we show are necessary and sufficient for domain generalization. Empirically, we show that TCRI is effective on both synthetic and real-world data. TCRI is competitive with baselines in average accuracy while outperforming them in worst-domain accuracy, indicating desired cross-domain stability.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。