Pseudo-label Learning with Calibrated Confidence Using an Energy-based Model
This work addresses the need for better confidence calibration in semi-supervised learning, which is crucial for improving pseudo-labeling accuracy in domains like image classification, though it is incremental in nature.
The paper tackles the problem of inaccurate confidence scores in pseudo-labeling for semi-supervised learning by proposing an energy-based pseudo-labeling (EBPL) algorithm that jointly trains a classifier and an energy-based model. The result shows that EBPL outperforms existing methods in semi-supervised image classification, with improved confidence calibration error and recognition accuracy.
In pseudo-labeling (PL), which is a type of semi-supervised learning, pseudo-labels are assigned based on the confidence scores provided by the classifier; therefore, accurate confidence is important for successful PL. In this study, we propose a PL algorithm based on an energy-based model (EBM), which is referred to as the energy-based PL (EBPL). In EBPL, a neural network-based classifier and an EBM are jointly trained by sharing their feature extraction parts. This approach enables the model to learn both the class decision boundary and input data distribution, enhancing confidence calibration during network training. The experimental results demonstrate that EBPL outperforms the existing PL method in semi-supervised image classification tasks, with superior confidence calibration error and recognition accuracy.