Comment on the subdifferential formula given in the article "Algorithmic Analysis and Statistical Estimation of SLOPE via Approximate Message Passing" - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2023

Comment on the subdifferential formula given in the article "Algorithmic Analysis and Statistical Estimation of SLOPE via Approximate Message Passing"

Résumé

Subdifferential formula for the sorted 1 norm attracted lot of attention recently to derive screening procedures for SLOPE [5, 7], clustering and sparsity properties for SLOPE [1, 10] and explicit expressions for the proximal operator of the sorted 1 norm [4, 11]. In this note, we discuss the formula for the subdifferential of the sorted 1 norm given in the article of Bu et al., [3]. We believe that expressing the sorted 1 norm as the maximum of linear functions, as suggested by the authors, is an appropriate approach to derive its subdifferential. However the formula provided in this article contains many flaws and we hope that authors would take into account this note in order to rewrite and prove their formula.
Fichier principal
Vignette du fichier
Note_IEEE.pdf (207.5 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04068605 , version 1 (14-04-2023)

Identifiants

  • HAL Id : hal-04068605 , version 1

Citer

Patrick J C Tardivel. Comment on the subdifferential formula given in the article "Algorithmic Analysis and Statistical Estimation of SLOPE via Approximate Message Passing". 2023. ⟨hal-04068605⟩
31 Consultations
29 Téléchargements

Partager

Gmail Facebook X LinkedIn More