What does example code do?

[zavalyshyn]

Hi,
I've been able to successfully run your rust/examples code (coordinator and clients + redis db).
In the description you mention that it's a basic federated learning algorithm but it's "network capable".
Could you pleas elaborate on this further?

What exactly is the algorithm the clients are running?
How do you aggregate the updates to the global model (if you do)?
What data the test example is running on? What do you train the model on?

Thanks

[Robert-Steiner]

Hi @zavalyshyn,
I'm sorry it took so long to reply.

In the description you mention that it's a basic federated learning algorithm but it's "network capable".
Could you pleas elaborate on this further?

The example is "network capable" in a sense that the client and the coordinator communicate over TCP. Since both the client and the coordinator are written in Rust, it would also be possible to have both communicate via in-memory channels.

What exactly is the algorithm the clients are running?
What data the test example is running on? What do you train the model on?

In terms of machine learning, the client doesn't run any algorithm, it just sends a dummy model every round it's chosen. The Rust example is intended more for testing the coordinator (stress test, etc.) and the PET protocol. If you are looking for a more real world example, I can recommend this Python example.

How do you aggregate the updates to the global model (if you do)?

In short: we use federated averaging in combination with homomorphic encryption. It is described in more detail in the whitepaper.

I hope is answers your questions, if not let me know.

[zavalyshyn]

@Robert-Steiner thanks a lot for your answers. They do answer all of my questions.
I was looking for some machine learning examples written in pure Rust and hoped to find it here. Seems like Python still prevails on that front.

Nevertheless, thanks for making Xaynet available to people like me and working on it. Keep up the good work!