This repository summaries the related work and advances on Federated learning focusing on incentives and attacks. every paper include would have a brief discription and pdf file.
[IN12] Understanding Federated Learning via Client-Level Influence Measurement.
This paper proposes Fed-Influence, to measure contribution based on sub-model parameter without retrainig as well as a estimation algorithm. It works well on both convex and non-convex loss functions and does not require the nal model to be optimal.