no code implementations • ICLR 2021 • Mary Phuong, Christoph H Lampert
We study the inductive bias of two-layer ReLU networks trained by gradient flow.
Inductive Bias