Arno Geimer, Beltran Fiz Pontiveros, Radu State
FedRandom is a novel technique that improves the stability and accuracy of participant contribution assessments in federated learning by generating more consistent samples, significantly reducing estimation errors.
Federated Learning allows multiple parties to collaboratively train machine learning models while keeping their data private. However, determining the fair contribution of each participant is challenging, especially when participants expect compensation or when identifying malicious actors. Existing methods for estimating contributions can be unstable, discouraging participation. FedRandom addresses this by treating contribution estimation as a statistical problem, allowing for more samples and thus more accurate evaluations. This approach leads to a more reliable assessment of each participant's role in training the model.
