On the inconsistency of separable losses for structured prediction - Traitement du Langage Parlé Access content directly
Conference Papers Year : 2023

On the inconsistency of separable losses for structured prediction

Caio Corro

Abstract

In this paper, we prove that separable negative log-likelihood losses for structured prediction are not necessarily Bayes consistent, or, in other words, minimizing these losses may not result in a model that predicts the most probable structure in the data distribution for a given input. This fact opens the question of whether these losses are well-adapted for structured prediction and, if so, why.
Fichier principal
Vignette du fichier
2023.eacl-main.109.pdf (315.7 Ko) Télécharger le fichier
Origin Publisher files allowed on an open archive

Dates and versions

hal-04394967 , version 1 (15-01-2024)

Identifiers

Cite

Caio Corro. On the inconsistency of separable losses for structured prediction. 17th Conference of the European Chapter of the Association for Computational Linguistics, May 2023, Dubrovnik, Croatia. pp.1491-1498, ⟨10.18653/v1/2023.eacl-main.109⟩. ⟨hal-04394967⟩
27 View
7 Download

Altmetric

Share

Gmail Mastodon Facebook X LinkedIn More