Tied-Augment: Controlling Representation Similarity Improves Data Augmentation

TL;DR

Tied-Augment introduces a simple loss term to control representation similarity under distortions, significantly enhancing data augmentation effectiveness across various training paradigms with fewer epochs.

Contribution

The paper proposes Tied-Augment, a general framework that improves data augmentation by controlling representation similarity, leading to faster and more effective training.

Findings

Tied-Augment outperforms baseline methods like RandAugment by 2.0% on ImageNet.

It enables data augmentation to improve generalization with fewer training epochs.

The method enhances semi-supervised and fine-tuning performance.

Abstract

Data augmentation methods have played an important role in the recent advance of deep learning models, and have become an indispensable component of state-of-the-art models in semi-supervised, self-supervised, and supervised training for vision. Despite incurring no additional latency at test time, data augmentation often requires more epochs of training to be effective. For example, even the simple flips-and-crops augmentation requires training for more than 5 epochs to improve performance, whereas RandAugment requires more than 90 epochs. We propose a general framework called Tied-Augment, which improves the efficacy of data augmentation in a wide range of applications by adding a simple term to the loss that can control the similarity of representations under distortions. Tied-Augment can improve state-of-the-art methods from data augmentation (e.g. RandAugment, mixup), optimization (e.g. SAM), and semi-supervised learning (e.g. FixMatch). For example, Tied-RandAugment can outperform RandAugment by 2.0% on ImageNet. Notably, using Tied-Augment, data augmentation can be made to improve generalization even when training for a few epochs and when fine-tuning. We open source our code at https://github.com/ekurtulus/tied-augment/tree/main.