3 ms·
In theory you could train locally and only send the combined gradient update to a central server, but I'm sure some people would argue that that's not fully pri
by WithinReason 2y ago
In theory you could train locally and only send the combined gradient update to a central server, but I'm sure some people would argue that that's not fully privacy preserving.