Objective: Build a simple federated learning model using a popular framework such as TensorFlow Federated or PySyft.
Description:
Implement a basic federated learning setup where multiple clients (agents) collaboratively train a machine learning model without sharing their local data.
Use a simple dataset, such as MNIST, and ensure that the model can aggregate updates from multiple clients.
The system should include basic local training, aggregation (e.g., Federated Averaging), and model evaluation.
Objectives:
-
Understand how federated learning works and implement a model that can run across multiple local devices.
-
Be able to aggregate updates from different devices in a secure and privacy-preserving way.
-
Evaluate the model’s accuracy after federated training and compare it to a centralized version of the same model.