Afficher la notice abrégée

dc.rights.licenseopenen_US
dc.contributor.authorAYYAR, Meghna P.
dc.contributor.authorBENOIS PINEAU, Jenny
IDREF: 074466992
dc.contributor.authorZEMMARI, Akka
IDREF: 157264491
hal.structure.identifierBordeaux population health [BPH]
dc.contributor.authorAMIEVA, Helene
dc.contributor.authorMIDDLETON, Laura
dc.contributor.editorCHETOUANI, Aladine
dc.contributor.editorBAILLER, Werner
dc.contributor.editorGURRIN, Cathal
dc.contributor.editorBENOIT, Alexandre
dc.date.accessioned2024-05-27T09:58:53Z
dc.date.available2024-05-27T09:58:53Z
dc.date.issued2023-12-30
dc.date.conference2023-09-20
dc.identifier.isbn979-8-4007-0912-8en_US
dc.identifier.urihttps://oskar-bordeaux.fr/handle/20.500.12278/200063
dc.description.abstractThe current paradigm of training deep neural networks relies on large, annotated and representative datasets. They assume a static world where the target domain does not change. However, in the real-world, data changes over time and is often available on the fly. Naive retraining on new data causes catastrophic forgetting and the network is unable to generalize on old data. Streaming learning is a type of incremental learning where networks learn sequentially and as soon as a sample is available from the data stream. Instead of training on every new sample, we propose an uncertainty based selection criteria to improve our previously proposed fast streaming learning method Move-to-Data (MTD), called Entropy-based MTD (EMTD). Besides, streaming learning methods have so far mostly used Convolutional Neural Networks (CNNs) but in recent times Vision Transformers (ViTs) have shown much better performances for many vision tasks. Therefore, we use ViT based Video Transformer to analyse MTD, EMTD and their gradient descent based "retargeting" steps. We have compared the performances of EMTD with MTD (w/wo retargeting) and a popular streaming learning method ExStream for the transformer. EMTD is able to outperform baseline MTD, and EMTD with retargeting achieves close results as ExStream and is ∼ 1.2 times faster.
dc.language.isoENen_US
dc.title.enEntropy-based Sampling for Streaming learning with Move-to-Data approach on Video
dc.typeCommunication dans un congrèsen_US
dc.identifier.doi10.1145/3617233.3617240en_US
dc.subject.halSciences du Vivant [q-bio]/Santé publique et épidémiologie
bordeaux.page21-27en_US
bordeaux.hal.laboratoriesBordeaux Population Health Research Center (BPH) - UMR 1219en_US
bordeaux.institutionUniversité de Bordeauxen_US
bordeaux.institutionINSERMen_US
bordeaux.conference.title20th International Conference on Content-based Multimedia Indexingen_US
bordeaux.countryfren_US
bordeaux.title.proceedingCBMI '23: Proceedings of the 20th International Conference on Content-based Multimedia Indexingen_US
bordeaux.teamACTIVE_BPHen_US
bordeaux.conference.cityOrléansen_US
hal.identifierhal-04589023
hal.version1
hal.date.transferred2024-05-28T23:04:30Z
hal.proceedingsouien_US
hal.conference.end2023-09-22
hal.popularnonen_US
hal.audienceInternationaleen_US
hal.exporttrue
workflow.import.sourcedissemin
dc.rights.ccPas de Licence CCen_US
bordeaux.COinSctx_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.date=2023-12-30&rft.spage=21-27&rft.epage=21-27&rft.au=AYYAR,%20Meghna%20P.&BENOIS%20PINEAU,%20Jenny&ZEMMARI,%20Akka&AMIEVA,%20Helene&MIDDLETON,%20Laura&rft.isbn=979-8-4007-0912-8&rft.genre=unknown


Fichier(s) constituant ce document

FichiersTailleFormatVue

Il n'y a pas de fichiers associés à ce document.

Ce document figure dans la(les) collection(s) suivante(s)

Afficher la notice abrégée