Ensembles depend on diversity for improved performance. Many ensemble training methods, therefore, attempt to optimize for diversity, which they almost always define in terms of differences in training set predictions. In this paper, however, we demonstrate the diversity of predictions on the training set does not necessarily imply diversity under mild covariate shift, which can harm generalization in practical settings. To address this issue, we introduce a new diversity metric and associated method of training ensembles of models that extrapolate differently on local patches of the data manifold. Across a variety of synthetic and real-world tasks, we find that our method improves generalization and diversity in qualitatively novel ways, especially under data limits and covariate shift.
@article{arxiv.1911.01291,
title = {Ensembles of Locally Independent Prediction Models},
author = {Andrew Slavin Ross and Weiwei Pan and Leo Anthony Celi and Finale Doshi-Velez},
journal= {arXiv preprint arXiv:1911.01291},
year = {2020}
}
Comments
This is an expansion of arXiv:1806.08716 with different applications and focus, accepted to AAAI 2020. Latest update clarifies a derivation