Skip to Main content Skip to Navigation
Conference papers

Oubli Catastrophique et Modèles Neuronaux de Recherche d'Information

Abstract : In this paper, we study in what extent neural ranking models catastrophically forget old knowledge acquired from previously observed domains after acquiring new knowledge (i.e.on new datasets), leading to performance decrease on those domains. Our experiments carried out on 4 datastes and 5 neural IR ranking models show that catastrophic forgetting is prevalentand that a lifelong learning strategy mitigates the problem. Using an explanatory approach builton a regression model, we also identify domain characteristics that can predict catastrophic for getting .Note: this paper is a summary of the ECIR 2021 paper (Lovón-Melgarejoet al., 2021).
Document type :
Conference papers
Complete list of metadata

https://hal.sorbonne-universite.fr/hal-03309985
Contributor : Laure Soulier Connect in order to contact the contributor
Submitted on : Friday, July 30, 2021 - 12:06:14 PM
Last modification on : Tuesday, October 19, 2021 - 2:24:25 PM

File

main (1).pdf
Files produced by the author(s)

Licence


Distributed under a Creative Commons Attribution 4.0 International License

Identifiers

  • HAL Id : hal-03309985, version 1

Citation

Jesus Lovon Melgarejo, Laure Soulier, Karen Pinel-Sauvagnat, Lynda Tamine. Oubli Catastrophique et Modèles Neuronaux de Recherche d'Information. 17ème conférence francophone en Recherche d’Information et Application (CORIA 2021), ARIA : Association Francophone de Recherche d’Information (RI) et Applications; Equipe MRIM du Laboratoire d’Informatique de Grenoble., Apr 2021, Grenoble (virtuel), France. pp.1-5. ⟨hal-03309985⟩

Share

Metrics

Record views

38

Files downloads

25