Level Data Augmentation

Level data augmentation techniques, encompassing modifications at the word, sentence, bag (a collection of instances), or even scene level, aim to enhance the robustness and performance of machine learning models, particularly in scenarios with limited labeled data or inherent biases. Current research focuses on developing augmentation strategies tailored to specific data types and model architectures, including those leveraging contrastive learning and semi-supervised learning frameworks, often combined with novel loss functions. These methods show promise in improving model generalization across various domains, such as natural language processing, computer vision (including whole slide image classification), and graph-based learning, leading to more accurate and fairer predictions in diverse applications.