RAPA-ConvNets: Modified Convolutional Networks for Accelerated Training on Architectures With Analog Arrays https://t.co/PT00WiBjB9 https://t.co/eLOGymL9ec
6,437 followers
9 followers
RT @MattiaRigotti: New paper out in @FrontNeurosci on how Analog Arrays will be able to accelerate Convolutional Neural Nets thanks to para…
88 followers
RT @MattiaRigotti: New paper out in @FrontNeurosci on how Analog Arrays will be able to accelerate Convolutional Neural Nets thanks to para…
565 followers
Thought analog HW is not good for ConvNets? Think twice and read this paper. Always thought the opposite? Good news for you! The authors also show more robustness to adversarial attacks. I have been waiting for a study like this for years! @aiCorTeX @theF