TY  - RPRT
TI  - Stochastic linear optimization never overfits with quadratically-bounded losses on general data
AU  - Matus Telgarsky
PY  - 2022
UR  - https://arxiv.org/abs/2202.06915
ID  - 2202.06915
ER  -