TinyFedTL: Federated Transfer Learning on Tiny Devices
This work solves privacy and hardware limitations for TinyML applications, though it appears incremental as it adapts existing federated and transfer learning concepts to a new platform.
The authors tackled the challenge of deploying machine learning on privacy-sensitive, resource-constrained tiny devices by implementing TinyFedTL, the first federated transfer learning system on a microcontroller, addressing memory and communication constraints.
TinyML has rose to popularity in an era where data is everywhere. However, the data that is in most demand is subject to strict privacy and security guarantees. In addition, the deployment of TinyML hardware in the real world has significant memory and communication constraints that traditional ML fails to address. In light of these challenges, we present TinyFedTL, the first implementation of federated transfer learning on a resource-constrained microcontroller.