[Federarated learning] we add a horizontal federated learning example

[482170765]

Feature Area

Many AI training use cases involves performance enhancement and security. Some times they are hard to balance. Distributed
training requires data to stay at the local site, while a summit model can be assembled at a virtual site. A federated learning fits such needs.

What feature would you like to see?

The architecture of the simulated federated learning is as follows. The clients will train on data in parallel in their own container and send model parameters information to the server. The federated learning works by a round table procedure where each client submits its training results to the server in each unit epoch. The server sums up the separate model parameters to form a new model. The new model is then passed on to each client residing on different local container. When they receive a response from server, clients will set the new model weights and proceed to the next federated learning round.

What is the use case or pain point?

many real world applications need such feature for AI security reasons.

Is there a workaround currently?

No

---

Love this idea? Give it a 👍.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[github-actions[bot]]

This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.