We consider training and testing on mixture distributions with different training and test proportions. We show that in many settings, and in some sense generically, distribution shift can be beneficial, and test performance can improve due to mismatched training proportions, even if the components are unrelated and with no transfer between components. In a variety of scenarios, we identify the optimal training proportions and the extent to which such distribution shift can be beneficial. We show how the same analysis applies also to a compositional setting with differing distribution of component "skills'' at training and test.

翻译：本文研究在训练和测试阶段使用不同混合比例分布的训练与测试问题。我们证明，在许多场景下，且从某种意义上具有普适性，分布偏移可能是有益的；即使各组成成分之间无关联且不存在成分间的知识迁移，不匹配的训练比例仍能提升测试性能。通过多种情境分析，我们确定了最优训练比例，并量化了此类分布偏移可能带来的性能增益程度。研究进一步表明，相同分析方法同样适用于训练与测试阶段成分'技能'分布存在差异的组合学习场景。