Robust deep active learning for data streams

Eiram Mahera Scheikh
Research Theme
R2 Prosilience & Robustness
Active Learning

The objective of this project is to make learning from evolving data streams label-efficient. The project integrates advanced deep learning techniques to model complex patterns and structures within continuously evolving data streams. The focus is on active learning strategies to intelligently identify a small subset of data for labeling to reduce the need for extensive labeled data and lower annotation costs. There is also an emphasis on robustness to ensure effective learning amidst the noise, outliers, and drifts inherent in streaming data. This research aspires to make contributions to domains characterized by a high cost of annotation and an evolving environment.