@j_foerst Not sure about EVs, but for GANs, in theory, game theory gives the solution: https://t.co/hXRTzFIrge The fact we have not been able to scale to high-d raises doubts on how good GANs really are... Happy to discuss implications for EV-adoption
1,644 followers
2,768 followers
RT @faoliehoek: @fhuszar @adjiboussodieng @peteskomoroch @AndrewYNg And, while it sounds interesting, there is plenty of ML fuzziness ("add…
1,644 followers
@fhuszar @adjiboussodieng @peteskomoroch @AndrewYNg And, while it sounds interesting, there is plenty of ML fuzziness ("added noise ... tractable approximations ... stabilizes the training procedure .... encourages ... to capture all the modes") Much less
2,376 followers
RT @faoliehoek: @fhuszar Great post - love the Escher comparison ;-) A problem remains: 'local Nash equilibria' may not possess the proper…