Tweeted By @Reza_Zadeh
Two excellent papers building on each other to show:
— Reza Zadeh (@Reza_Zadeh) November 16, 2018
1) The loss of two-layer neural networks can be optimized to zero in polynomial time using gradient descent: https://t.co/pkvzNMH7hX
2) Same for ResNet, teasing out benefits of residual connections: https://t.co/ZfS1eW4us8 pic.twitter.com/5Tb5m0fia8