Effective Augmentation
Effective data augmentation aims to improve the performance and robustness of machine learning models by artificially expanding training datasets. Current research focuses on developing augmentation strategies tailored to specific data types (e.g., time series, graphs, images) and model architectures (e.g., GANs, GNNs, contrastive learning models), often employing techniques like mixup, cutout, and generative models to create diverse synthetic data. These advancements are crucial for addressing challenges like data scarcity, improving model generalization, and enhancing efficiency in various applications, including medical imaging, remote sensing, and natural language processing. The development of automated augmentation strategies, guided by dataset characteristics or model performance, is a significant area of ongoing investigation.
Papers
Learning Better with Less: Effective Augmentation for Sample-Efficient Visual Reinforcement Learning
Guozheng Ma, Linrui Zhang, Haoyu Wang, Lu Li, Zilin Wang, Zhen Wang, Li Shen, Xueqian Wang, Dacheng Tao
Dynamic Data Augmentation via MCTS for Prostate MRI Segmentation
Xinyue Xu, Yuhan Hsi, Haonan Wang, Xiaomeng Li