Repetitive Reprediction Deep Decipher for Semi-Supervised Learning

TL;DR

This paper introduces a theoretically grounded semi-supervised learning framework called deep decipher (D2), which uses a novel repetitive reprediction strategy to improve pseudo-label quality, demonstrating superior performance on ImageNet.

Contribution

The paper provides a theoretical link between network predictions and pseudo-labels, and proposes the R2 strategy to enhance pseudo-label reliability in semi-supervised learning.

Findings

R2-D2 outperforms state-of-the-art methods by 5% on ImageNet.

Theoretical support for using network predictions as pseudo-labels.

Repetitive reprediction improves pseudo-label certainty.

Abstract

Most recent semi-supervised deep learning (deep SSL) methods used a similar paradigm: use network predictions to update pseudo-labels and use pseudo-labels to update network parameters iteratively. However, they lack theoretical support and cannot explain why predictions are good candidates for pseudo-labels. In this paper, we propose a principled end-to-end framework named deep decipher (D2) for SSL. Within the D2 framework, we prove that pseudo-labels are related to network predictions by an exponential link function, which gives a theoretical support for using predictions as pseudo-labels. Furthermore, we demonstrate that updating pseudo-labels by network predictions will make them uncertain. To mitigate this problem, we propose a training strategy called repetitive reprediction (R2). Finally, the proposed R2-D2 method is tested on the large-scale ImageNet dataset and outperforms state-of-the-art methods by 5 percentage points.