Latent Class-Conditional Noise Model

TL;DR

This paper introduces a Bayesian latent class-conditional noise model (LCCN) that effectively learns noise transitions in label-noise scenarios, improving robustness and stability over previous methods, and extends to various noisy label settings.

Contribution

The paper proposes a novel LCCN framework that constrains noise transition learning within a Dirichlet space, avoiding local minima issues and enabling extensions to open-set, semi-supervised, and cross-model learning.

Findings

Outperforms state-of-the-art methods on noisy label benchmarks.

Provides stable and efficient noise transition estimation.

Extends to open-set and semi-supervised noisy label scenarios.

Abstract

Learning with noisy labels has become imperative in the Big Data era, which saves expensive human labors on accurate annotations. Previous noise-transition-based methods have achieved theoretically-grounded performance under the Class-Conditional Noise model (CCN). However, these approaches builds upon an ideal but impractical anchor set available to pre-estimate the noise transition. Even though subsequent works adapt the estimation as a neural layer, the ill-posed stochastic learning of its parameters in back-propagation easily falls into undesired local minimums. We solve this problem by introducing a Latent Class-Conditional Noise model (LCCN) to parameterize the noise transition under a Bayesian framework. By projecting the noise transition into the Dirichlet space, the learning is constrained on a simplex characterized by the complete dataset, instead of some ad-hoc parametric space wrapped by the neural layer. We then deduce a dynamic label regression method for LCCN, whose Gibbs sampler allows us efficiently infer the latent true labels to train the classifier and to model the noise. Our approach safeguards the stable update of the noise transition, which avoids previous arbitrarily tuning from a mini-batch of samples. We further generalize LCCN to different counterparts compatible with open-set noisy labels, semi-supervised learning as well as cross-model training. A range of experiments demonstrate the advantages of LCCN and its variants over the current state-of-the-art methods.