Bridging the Gap: Learning Pace Synchronization for Open-World Semi-Supervised Learning

TL;DR

This paper introduces a novel approach for open-world semi-supervised learning that balances learning between seen and novel categories, significantly improving accuracy on benchmark datasets by using adaptive loss and pseudo-label clustering.

Contribution

The paper proposes the adaptive synchronizing marginal loss and pseudo-label contrastive clustering to address class imbalance and semantic understanding in open-world semi-supervised learning.

Findings

Achieves 3% average accuracy increase on ImageNet.

Balances learning pace between seen and novel classes effectively.

Fine-tuning pre-trained models further boosts performance.

Abstract

In open-world semi-supervised learning, a machine learning model is tasked with uncovering novel categories from unlabeled data while maintaining performance on seen categories from labeled data. The central challenge is the substantial learning gap between seen and novel categories, as the model learns the former faster due to accurate supervisory information. Moreover, capturing the semantics of unlabeled novel category samples is also challenging due to the missing label information. To address the above issues, we introduce 1) the adaptive synchronizing marginal loss which imposes class-specific negative margins to alleviate the model bias towards seen classes, and 2) the pseudo-label contrastive clustering which exploits pseudo-labels predicted by the model to group unlabeled data from the same category together in the output space. Extensive experiments on benchmark datasets demonstrate that previous approaches may significantly hinder novel class learning, whereas our method strikingly balances the learning pace between seen and novel classes, achieving a remarkable 3% average accuracy increase on the ImageNet dataset. Importantly, we find that fine-tuning the self-supervised pre-trained model significantly boosts the performance, which is overlooked in prior literature. Our code is available at https://github.com/yebo0216best/LPS-main.