TY  - RPRT
TI  - Gradient Descent can Learn Less Over-parameterized Two-layer Neural Networks on Classification Problems
AU  - Atsushi Nitanda
AU  - Geoffrey Chinot
AU  - Taiji Suzuki
PY  - 2020
UR  - https://arxiv.org/abs/1905.09870
ID  - 1905.09870
ER  -