Federated Learning

YOUTUBE 1YbPmkChcbo TensorFlow Federated (TFF): Machine Learning on Decentralized Data (TF Dev Summit ‘19)

Current model. Training data is living in the cloud. We distribute the model to devices.

As a social network one of the crucial design differences between Federated Wiki and Facebook. Between Ward Cunningham and Mark Zuckerberg is respect to feedback. Facebook closes a social feedback loop to deliberately generate strong emotions for ad-tech. Wiki does not. Without paying lots of attention this leaves wiki bereft of network feedback.

Wiki learned from Wikipedia to stop heated political fights not foster then with feedback. David Bovill has spoken about this lack of feedback to understand the sentiment inside the Federation.

Tensorflow Federated Learning uses 'differential privacy' within its machine learning. I've heard directly from Emily Glanz is a software engineer at Google! working on TensorFlow Federated about its construct and believe Wiki would benefit from learning more by inviting the TFFL team to share ideas at Ward's weekly online meeting.