Fetching the paper…

Stochastic linear optimization never overfits with quadratically-bounded losses on general data · Around