TY  - RPRT
TI  - Deep learning generalizes because the parameter-function map is biased towards simple functions
AU  - Guillermo Valle-Pérez
AU  - Chico Q. Camargo
AU  - Ard A. Louis
PY  - 2019
UR  - https://arxiv.org/abs/1805.08522
ID  - 1805.08522
ER  -