2 ms·
Private Multi-Party Machine Learning
- webmaven 10y agoSlides from the scheduled presentations[0], as well as most of the accepted papers[1], are available. This is some seriously mind-bending stuff. Google is represented by a couple of papers, but I was really hoping that the privacy-preserving techniques they used in training the network for Inbox Smart Replies would be explained. No such luck. [0] https://pmpml.github.io/PMPML16/#schedule https://pmpml.github.io/PMPML16/#schedule [1] https://pmpml.github.io/PMPML16/#papers https://pmpml.github.io/PMPML16/#papers
- jhoechtl 10y agoI wonder why nobody of these guys has been there http://www.enigma.co http://www.enigma.co