A visual Introduction to Federated or Collaborative Learning

[@alx8439]

I knew why PCs running Windows are constantly eating CPU - it's just Microsoft using these machines for their own ML needs, utilizing this federated learning concept

[@650AILab]

Almost all very big organization i.e. Google, Facebook, Microsoft, LinkedIn they all are utilizing federated learning to utilize the edge devices smartly.

[@alx8439]

@@650AILab I mean - they are not only utilizing their own devices, but also user devices of those who use their products :)

[@aneesharma3240]

My questions are :
1. The global model is sent to the devices, the devices train on their training data and sends back the weights/model parameters to the server. The server aggregates and updates the global model and sents back to the devices for further training. When is the evaluation done ?
2. The devices may have different different data distributions, a model may perform well in one device and might perform poorly in another device. How can it be handled?

[@sachugowda]

1) You can do evaluation in both client and server
2) The solution is Personalized FL. (We can build Personalized FL models to work better for individuals clients)

[@zahrazamir7896]

Aoa sir, can you tell which parameters or data is used to design the model initially at the server.

[@user-mc3th9xk5v]

Thanks sir

[@RaushanKumarSingh-gi6fw]

WHICH SOFTWARE U USED TO CREATE
DIAGRAMS