A arxiv.org Every model learned by gradient descent is approximately a kernel machine (2020) 176 points Anon84 2 years ago 136 comments
