Tf.js + federated learning sample

I was wondering if there’s an openly available sample or demo of:

  • A model that trains and inferences in the browser
  • And updates a global model using federated learning

I was thinking of something like hand gestures model which retrains locally in the browser based on new user actions and model weights are updated on a central model.

Am I thinking it properly? Is there some demo/sample I can use as starting point? Is this even feasible?


I don’t think there are so much updates since 2019:

1 Like

Take a look at:

1 Like

Oooh this seems very interesting. Thanks for sharing @Bhack - somehow missed this on my radar! Are you working on this?

It Is a team from Machine Learning and Optimization Laboratory ‐ EPFL

1 Like

Disco 1.0 Is released: