Exécution de notre outil de synthèse sur un nouvel article
Vérifiez l’article ci-dessous, puis lancez sa synthèse.
sparse-lm: Sparse linear regression models in Python
Luis Barroso-Luque (Lawrence Berkeley National Laboratory, University of California, Berkeley), Fengyu Xie (Lawrence Berkeley National Laboratory, University of California, Berkeley)
Résumé : Sparse linear regression models are a powerful tool for capturing linear relationships in high dimensional spaces.Sparse models have only a small number of nonzero parameters -even if the number of covariates used in estimation is large-as a result they can be easier to fit and interpret compared to dense models (Hastie et al., 2015).Regression objectives resulting in sparse linear models such as the Lasso (Tibshirani, 2018;Zou, 2006) and Best Subset Selection (Hocking & Leslie, 1967) have been widely used in a variety of fields.However, many regression problems involve covariates that have a natural underlying structure, such as group or hierarchical relationships, that can be further leveraged to obtain improved model performance and interpretability.Such structured regression problems occur in a wide range of fields including genomics (Chen & Wang, 2021), bioinformatics (Ma et al., 2007), medicine (Kim et al., 2012), econometrics (Athey & Imbens, 2017), chemistry (Gu et al., 2018), and materials science (Leong & Tan, 2019).Several generalizations of the Lasso (Friedman et al., 2010;Simon et al., 2013;Wang & Tian, 2019;Yuan & Lin, 2006) and Best Subset Selection (Bertsimas et al., 2016; Bertsimas & King, 2016) have been developed to effectively exploit additional structure in linear regression.The sparse-lm Python package provides a flexible, comprehensive, and user-friendly implementation of (structured) sparse linear regression models, which allows researchers to easily experiment and choose the best regression model for their specific problem.