Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2021
|
| Soggetti: | |
| Accesso online: | https://arxiv.org/abs/2110.11074 |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866910272608796672 |
|---|---|
| author | Yang, Archer Y. Zhao, Yue Lian, Yi Gu, Yuwen Fan, Jun |
| author_facet | Yang, Archer Y. Zhao, Yue Lian, Yi Gu, Yuwen Fan, Jun |
| contents | Many statistics problems are formulated within an estimating equation framework instead of a minimization framework. However, the regularized estimating equations (REE) have been much less extensively studies than regularized minimization problems. In this paper, we study an improved regularized estimating equation formulation and explore its subsequent equivalences in terms of (1) fixed-point problem specified via the proximal operator of the corresponding regularizer, and (2) generalized variational inequality problems. Such equivalences hold under general conditions and accommodate nonconvex regularizers. Moreover, these equivalences open up new possibilities in theoretical analysis and computational algorithms when studying the REE. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2110_11074 |
| institution | arXiv |
| publishDate | 2021 |
| record_format | arxiv |
| spellingShingle | A Unified Framework for Regularized Estimating Equations via Fixed-Point and Variational Inequality Problems Yang, Archer Y. Zhao, Yue Lian, Yi Gu, Yuwen Fan, Jun Methodology Many statistics problems are formulated within an estimating equation framework instead of a minimization framework. However, the regularized estimating equations (REE) have been much less extensively studies than regularized minimization problems. In this paper, we study an improved regularized estimating equation formulation and explore its subsequent equivalences in terms of (1) fixed-point problem specified via the proximal operator of the corresponding regularizer, and (2) generalized variational inequality problems. Such equivalences hold under general conditions and accommodate nonconvex regularizers. Moreover, these equivalences open up new possibilities in theoretical analysis and computational algorithms when studying the REE. |
| title | A Unified Framework for Regularized Estimating Equations via Fixed-Point and Variational Inequality Problems |
| topic | Methodology |
| url | https://arxiv.org/abs/2110.11074 |