Abstract:
This BTech project aims to explore the use of two pivotal concepts in machinelearning: coresets and federated learning. Leveraging insights from the AIsummer school hosted by the esteemed advisors in these domains, this projectseeks to innovate novel methodologies at the intersection of these cutting-edgeareas. Coresets play a pivotal role in reducing dataset sizes while maintainingmodel accuracy, which aligns well with federated learning's focus ondecentralized, collaborative model training across diverse data sources. Theresearch will focus on devising efficient coreset construction techniques suitablefor federated learning scenarios. Ultimately, the goal is to be able to present thiswork as a research paper, contributing new perspectives and methodologies tothese emerging fields.