Negative Pseudo Label

Negative pseudo-labeling is a technique in semi-supervised learning that leverages unlabeled data by assigning negative labels to instances, improving model accuracy and robustness. Current research focuses on mitigating biases and errors inherent in pseudo-label generation, employing methods like contrastive learning, multi-view learning, and adaptive thresholding within various model architectures including neural networks and graph neural networks. This approach is particularly valuable in scenarios with limited labeled data, impacting fields like image classification, text classification, and multi-label learning by enhancing model performance and efficiency.