TY  - RPRT
TI  - L4: Practical loss-based stepsize adaptation for deep learning
AU  - Michal Rolinek
AU  - Georg Martius
PY  - 2018
UR  - https://arxiv.org/abs/1802.05074
ID  - 1802.05074
ER  -