@misc{indiciaef4881440646e,
  title = {The Convex Geometry of Backpropagation: Neural Network Gradient Flows Converge to Extreme Points of the Dual Convex Program},
  author = {Yifei Wang and Mert Pilanci},
  year = {2021},
  url = {https://arxiv.org/abs/2110.06488},
  note = {Source identifier: 2110.06488}
}