Near-optimal learning with average Hölder smoothness | Read Paper on Bytez