TMLC-Net: Transferable Meta Label Correction for Noisy Label Learning

2025-02-11

AI Summary

Noisy labels in real-world data severely hinder the robust deployment of deep learning models, while existing meta-learning approaches suffer from poor generalizability and heavy reliance on task-specific customization. To address this, we propose TMLC-Net, a transferable meta-learning label correction framework that establishes, for the first time, a universal, retraining-free label correction paradigm across datasets and model architectures. Its core innovations include: (1) a normalized noise-aware module that adaptively models sample-wise confidence; (2) an RNN-based temporal encoding module that captures dynamic training evolution patterns; and (3) a sub-class distribution decoding module enabling fine-grained modeling of noise structure. Extensive experiments under diverse noise types and intensities demonstrate that TMLC-Net significantly outperforms state-of-the-art methods. Crucially, it exhibits strong cross-dataset and cross-noise-type generalization, achieving substantial improvements in both classification accuracy and robustness.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Natural Language Processing: Safety and Robustness
• Multiagent Systems: Multiagent Learning

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

The prevalence of noisy labels in real-world datasets poses a significant impediment to the effective deployment of deep learning models. While meta-learning strategies have emerged as a promising approach for addressing this challenge, existing methods often suffer from limited transferability and task-specific designs. This paper introduces TMLC-Net, a novel Transferable Meta-Learner for Correcting Noisy Labels, designed to overcome these limitations. TMLC-Net learns a general-purpose label correction strategy that can be readily applied across diverse datasets and model architectures without requiring extensive retraining or fine-tuning. Our approach integrates three core components: (1) Normalized Noise Perception, which captures and normalizes training dynamics to handle distribution shifts; (2) Time-Series Encoding, which models the temporal evolution of sample statistics using a recurrent neural network; and (3) Subclass Decoding, which predicts a corrected label distribution based on the learned representations. We conduct extensive experiments on benchmark datasets with various noise types and levels, demonstrating that TMLC-Net consistently outperforms state-of-the-art methods in terms of both accuracy and robustness to label noise. Furthermore, we analyze the transferability of TMLC-Net, showcasing its adaptability to new datasets and noise conditions, and establishing its potential as a broadly applicable solution for robust deep learning in noisy environments.

Problem

• Improves noisy label correction in deep learning
• Enhances transferability across datasets and models
• Introduces adaptable strategies for various noise types

Innovation

• Transferable Meta-Learner
• Normalized Noise Perception
• Time-Series Encoding