Distributionally Robust Safe Sample Elimination under Covariate Shift

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Hanada, Hiroyuki, Aoyama, Tatsuya, Akahane, Satoshi, Tanaka, Tomonari, Okura, Yoshito, Inatsu, Yu, Hashimoto, Noriaki, Takeno, Shion, Murayama, Taro, Lee, Hanju, Kojima, Shinya, Takeuchi, Ichiro
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866913577589276672
author Hanada, Hiroyuki
Aoyama, Tatsuya
Akahane, Satoshi
Tanaka, Tomonari
Okura, Yoshito
Inatsu, Yu
Hashimoto, Noriaki
Takeno, Shion
Murayama, Taro
Lee, Hanju
Kojima, Shinya
Takeuchi, Ichiro
author_facet Hanada, Hiroyuki
Aoyama, Tatsuya
Akahane, Satoshi
Tanaka, Tomonari
Okura, Yoshito
Inatsu, Yu
Hashimoto, Noriaki
Takeno, Shion
Murayama, Taro
Lee, Hanju
Kojima, Shinya
Takeuchi, Ichiro
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2406_05964
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Distributionally Robust Safe Sample Elimination under Covariate Shift
Hanada, Hiroyuki
Aoyama, Tatsuya
Akahane, Satoshi
Tanaka, Tomonari
Okura, Yoshito
Inatsu, Yu
Hashimoto, Noriaki
Takeno, Shion
Murayama, Taro
Lee, Hanju
Kojima, Shinya
Takeuchi, Ichiro
Machine Learning
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.
title Distributionally Robust Safe Sample Elimination under Covariate Shift
topic Machine Learning
url https://arxiv.org/abs/2406.05964