Techniques to condense large datasets into smaller, informative subsets. Facilitates efficient training and data sharing while preserving model performance.
