What does "Federated Split Learning" mean?
Table of Contents
Federated Split Learning is a method used to train smart systems that recognize human activities, like walking or running, while keeping personal data private. It combines data from sensors, like accelerometers and gyroscopes, found in many devices. This way, the system can learn to identify actions more accurately.
How It Works
In this approach, the learning process happens in two parts. One part runs on a powerful server, and the other runs on your device, like a phone or a smartwatch. The server does the heavy lifting, while your device only handles a small part. This setup allows the system to use personal features without sending all your data to the server.
Benefits
Using Federated Split Learning can lead to better performance compared to standard methods. It not only improves the accuracy of recognizing activities but also reduces the time needed to send information back and forth between your device and the server.
Privacy Considerations
While this method helps in keeping data private, there are still some risks. When the system shares details during training, there's a chance some personal information could be exposed. However, adding a small layer of privacy protection can help lower this risk without hurting the system's ability to learn.