Enregistré dans:
Détails bibliographiques
Auteur principal: Barbalau, Antonio
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2411.15272
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866915832483807232
author Barbalau, Antonio
author_facet Barbalau, Antonio
contents In subpopulation shift scenarios, a Curriculum Learning (CL) approach would only serve to imprint the model weights, early on, with the easily learnable spurious correlations featured. To the best of our knowledge, none of the current state-of-the-art subpopulation shift approaches employ any kind of curriculum. To overcome this, we design a CL approach aimed at initializing the model weights in an unbiased vantage point in the hypothesis space which sabotages easy convergence towards biased hypotheses during the final optimization based on the entirety of the available data. We hereby propose a Curriculum-enhanced Group Distributionally Robust Optimization (CeGDRO) approach, which prioritizes the hardest bias-confirming samples and the easiest bias-conflicting samples, leveraging GroupDRO to balance the initial discrepancy in terms of difficulty. We benchmark our proposed method against the most popular subpopulation shift datasets, showing an increase over the state-of-the-art results across all scenarios, up to 6.2% on Waterbirds.
format Preprint
id arxiv_https___arxiv_org_abs_2411_15272
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Curriculum-enhanced GroupDRO: Challenging the Norm of Avoiding Curriculum Learning in Subpopulation Shift Setups
Barbalau, Antonio
Machine Learning
Artificial Intelligence
In subpopulation shift scenarios, a Curriculum Learning (CL) approach would only serve to imprint the model weights, early on, with the easily learnable spurious correlations featured. To the best of our knowledge, none of the current state-of-the-art subpopulation shift approaches employ any kind of curriculum. To overcome this, we design a CL approach aimed at initializing the model weights in an unbiased vantage point in the hypothesis space which sabotages easy convergence towards biased hypotheses during the final optimization based on the entirety of the available data. We hereby propose a Curriculum-enhanced Group Distributionally Robust Optimization (CeGDRO) approach, which prioritizes the hardest bias-confirming samples and the easiest bias-conflicting samples, leveraging GroupDRO to balance the initial discrepancy in terms of difficulty. We benchmark our proposed method against the most popular subpopulation shift datasets, showing an increase over the state-of-the-art results across all scenarios, up to 6.2% on Waterbirds.
title Curriculum-enhanced GroupDRO: Challenging the Norm of Avoiding Curriculum Learning in Subpopulation Shift Setups
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2411.15272