DCAST: Diverse Class-Aware Self-Training Mitigates Selection Bias for Fairer Learning

Research output: Working paper/Preprint › Preprint

Keywords

This output contributes to the following UN Sustainable Development Goals (SDGs)

@techreport{32580535a1814c5486df268bfd18a65e,

title = "DCAST: Diverse Class-Aware Self-Training Mitigates Selection Bias for Fairer Learning",

keywords = "selection bias, semi-supervised learning, self-training, distribution shift, diversity",

author = "Y.I. Tepeli and {P. Gon{\c c}alves}, Joana",

year = "2024",

month = oct,

day = "8",

doi = "10.48550/arXiv.2409.20126",

language = "English",

type = "WorkingPaper",

}

TY - UNPB

T1 - DCAST: Diverse Class-Aware Self-Training Mitigates Selection Bias for Fairer Learning

AU - Tepeli, Y.I.

AU - P. Gonçalves, Joana

PY - 2024/10/8

Y1 - 2024/10/8

KW - selection bias

KW - semi-supervised learning

KW - self-training

KW - distribution shift

KW - diversity

U2 - 10.48550/arXiv.2409.20126

DO - 10.48550/arXiv.2409.20126

M3 - Preprint

BT - DCAST: Diverse Class-Aware Self-Training Mitigates Selection Bias for Fairer Learning

ER -