Distributionally Robust Safe Sample Elimination under Covariate Shift
This addresses cost reduction for machine learning practitioners deploying customized models in varied environments, but it is incremental as it builds on existing techniques like distributionally robust optimization and safe sample screening.
The paper tackles the problem of reducing storage and training costs when training multiple models across different data distributions by proposing the DRSSS method, which combines distributionally robust optimization and safe sample screening to ensure models trained on a reduced dataset perform identically to those on the full dataset for all environments, with experiments demonstrating its effectiveness under covariate shift.
We consider a machine learning setup where one training dataset is used to train multiple models across slightly different data distributions. This occurs when customized models are needed for various deployment environments. To reduce storage and training costs, we propose the DRSSS method, which combines distributionally robust (DR) optimization and safe sample screening (SSS). The key benefit of this method is that models trained on the reduced dataset will perform the same as those trained on the full dataset for all possible different environments. In this paper, we focus on covariate shift as a type of data distribution change and demonstrate the effectiveness of our method through experiments.