📅 Date: Friday, November 14, 2025
🕚 Time: 11:00 – 12:00
Location: Heinzel Seminar Room, Building West
Speaker
Nikola Konstantinov (INSAIT)
Title
Incentivizing Participation and Collaboration in Federated Learning
Abstract
Collaborative and federated learning (FL) techniques have the potential to enable training powerful machine learning models from distributed data. However, in many cases, potential participants in such collaborative schemes have incentives beyond model accuracy. For example, they may be competitors on downstream tasks, be concerned about data privacy, or have differing data distributions.
These factors can lead participants to opt out of training, obfuscate messages, or even interfere with the process—potentially undermining the benefits of collaboration.
In this talk, Nikola Konstantinov will present recent results on incentivizing participation and honesty in federated learning. He will cover theoretical models for rational data-sharing decision-making in contexts involving market competition, privacy concerns, and data heterogeneity. The talk will also introduce federated learning protocols that provably incentivize honesty during training, even when participants’ incentives conflict.