ARA: Ensuring Fairness via Federated Learning Beyond Consensus

Faculty

Dinesh Manocha, Amrit Bedi

Funding Agency

Amazon Research Awards

Year

2022

Descriptions

"Ensuring Fairness via Federated Learning Beyond Consensus" is $50K in unrestricted funding from Amazon Research Awards. This is an AI/ML project.

Federated learning is a decentralized way to unlock information that feeds new AI applications and trains existing AI models without individual users’ data being seen or collected. The actual data never leave an individual mobile phone, laptop, or private server.

Federated learning is fast becoming the standard that complies with new regulations for handling and storing private data. It also offers a way to tap raw data streaming from sensors on satellites, bridges, machines, and a growing number of smart devices.

The new technique enables mobile phones to collaboratively learn a shared prediction model while keeping the training data on the device, decoupling the ability to do machine learning from the need to store data in the cloud. Federated learning goes beyond the use of local models that make predictions on mobile devices by bringing model training to the device as well.

A device downloads the current model, improves it by learning from data on the phone, then summarizes the changes as a small focused update. Only this update is sent to the cloud, using encrypted communication. It is then averaged with other user updates to improve the shared model. Training data remains on the device itself; no individual updates are stored in the cloud.

Federated learning produces smarter models, lower latency, and less power consumption, all while ensuring privacy. In addition to providing an update to the shared model, the improved model can also be used immediately on the device for a more personalized experience.