Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2022
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2211.10054 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909208315691008 |
|---|---|
| author | Liao, Yufan Wu, Qi Wu, Zhaodi Yan, Xing |
| author_facet | Liao, Yufan Wu, Qi Wu, Zhaodi Yan, Xing |
| contents | Invariant learning methods, aimed at identifying a consistent predictor across multiple environments, are gaining prominence in out-of-distribution (OOD) generalization. Yet, when environments aren't inherent in the data, practitioners must define them manually. This environment partitioning--algorithmically segmenting the training dataset into environments--crucially affects invariant learning's efficacy but remains underdiscussed. Proper environment partitioning could broaden the applicability of invariant learning and enhance its performance. In this paper, we suggest partitioning the dataset into several environments by isolating low-correlation data subsets. Through experiments with synthetic and real data, our Decorr method demonstrates superior performance in combination with invariant learning. Decorr mitigates the issue of spurious correlations, aids in identifying stable predictors, and broadens the applicability of invariant learning methods. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2211_10054 |
| institution | arXiv |
| publishDate | 2022 |
| record_format | arxiv |
| spellingShingle | Decorr: Environment Partitioning for Invariant Learning and OOD Generalization Liao, Yufan Wu, Qi Wu, Zhaodi Yan, Xing Machine Learning Invariant learning methods, aimed at identifying a consistent predictor across multiple environments, are gaining prominence in out-of-distribution (OOD) generalization. Yet, when environments aren't inherent in the data, practitioners must define them manually. This environment partitioning--algorithmically segmenting the training dataset into environments--crucially affects invariant learning's efficacy but remains underdiscussed. Proper environment partitioning could broaden the applicability of invariant learning and enhance its performance. In this paper, we suggest partitioning the dataset into several environments by isolating low-correlation data subsets. Through experiments with synthetic and real data, our Decorr method demonstrates superior performance in combination with invariant learning. Decorr mitigates the issue of spurious correlations, aids in identifying stable predictors, and broadens the applicability of invariant learning methods. |
| title | Decorr: Environment Partitioning for Invariant Learning and OOD Generalization |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2211.10054 |