4 ms·
"Homomorphic encryption enables the aggregation of these distributed model pieces while they are encrypted, allowing for federated learning without centralizing
by hunterpayne 1y ago
"Homomorphic encryption enables the aggregation of these distributed model pieces while they are encrypted, allowing for federated learning without centralizing data."
A bigger hand wave has never been done I think. Homomorphic encryption increases computational load several fold. And I'm not aware of anyone trying to use this (very interesting) technology for much of anything, let alone GPU ML algorithms.
- rafale 1y agoDidn't a hedge fund publish data encrypted with homomorphic encryption to run an open competition to see how can build the best trading AI. The encryption allow them to keep the sensitive data private.
- williamtrask 1y ago(OP here) Homomorphic addition (e.g. aggregation) is very performant, including for the Federated Averaging algorithm used in Federated Learning. Not hand-waivey.
- Yoric 1y agoDoesn't Zama have an homomorphic machine learning product?
- williamtrask 1y agoYeah Zama's stuff is great.