Contrastive Representations for Label Noise Require Fine-Tuning

TL;DR

This paper demonstrates that fine-tuning contrastive representations significantly improves label noise robustness in classification tasks, outperforming frozen representations and recent state-of-the-art methods across various noise types and levels.

Contribution

It shows that fine-tuning contrastive representations is essential for noise-robust classification, providing a new benchmark and insights into the importance of representation adaptation.

Findings

Fine-tuning contrastive representations outperforms frozen ones under label noise.

The proposed approach achieves state-of-the-art results across multiple noise types.

Results are stable across different noise levels.

Abstract

In this paper we show that the combination of a Contrastive representation with a label noise-robust classification head requires fine-tuning the representation in order to achieve state-of-the-art performances. Since fine-tuned representations are shown to outperform frozen ones, one can conclude that noise-robust classification heads are indeed able to promote meaningful representations if provided with a suitable starting point. Experiments are conducted to draw a comprehensive picture of performances by featuring six methods and nine noise instances of three different kinds (none, symmetric, and asymmetric). In presence of noise the experiments show that fine tuning of Contrastive representation allows the six methods to achieve better results than end-to-end learning and represent a new reference compare to the recent state of art. Results are also remarkable stable versus the noise level.