Saved in:
Bibliographic Details
Main Authors: Bankes, William, Hughes, George, Bogunovic, Ilija, Wang, Zi
Format: Preprint
Published: 2023
Subjects:
Online Access:https://arxiv.org/abs/2312.00486
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916493422231552
author Bankes, William
Hughes, George
Bogunovic, Ilija
Wang, Zi
author_facet Bankes, William
Hughes, George
Bogunovic, Ilija
Wang, Zi
contents Modern machine learning models are becoming increasingly expensive to train for real-world image and text classification tasks, where massive web-scale data is collected in a streaming fashion. To reduce the training cost, online batch selection techniques have been developed to choose the most informative datapoints. However, these techniques can suffer from poor worst-class generalization performance due to class imbalance and distributional shifts. This work introduces REDUCR, a robust and efficient data downsampling method that uses class priority reweighting. REDUCR reduces the training data while preserving worst-class generalization performance. REDUCR assigns priority weights to datapoints in a class-aware manner using an online learning algorithm. We demonstrate the data efficiency and robust performance of REDUCR on vision and text classification tasks. On web-scraped datasets with imbalanced class distributions, REDUCR significantly improves worst-class test accuracy (and average accuracy), surpassing state-of-the-art methods by around 15%.
format Preprint
id arxiv_https___arxiv_org_abs_2312_00486
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle REDUCR: Robust Data Downsampling Using Class Priority Reweighting
Bankes, William
Hughes, George
Bogunovic, Ilija
Wang, Zi
Machine Learning
Modern machine learning models are becoming increasingly expensive to train for real-world image and text classification tasks, where massive web-scale data is collected in a streaming fashion. To reduce the training cost, online batch selection techniques have been developed to choose the most informative datapoints. However, these techniques can suffer from poor worst-class generalization performance due to class imbalance and distributional shifts. This work introduces REDUCR, a robust and efficient data downsampling method that uses class priority reweighting. REDUCR reduces the training data while preserving worst-class generalization performance. REDUCR assigns priority weights to datapoints in a class-aware manner using an online learning algorithm. We demonstrate the data efficiency and robust performance of REDUCR on vision and text classification tasks. On web-scraped datasets with imbalanced class distributions, REDUCR significantly improves worst-class test accuracy (and average accuracy), surpassing state-of-the-art methods by around 15%.
title REDUCR: Robust Data Downsampling Using Class Priority Reweighting
topic Machine Learning
url https://arxiv.org/abs/2312.00486