Found 1 document(s)
Page 1 of 1
-
Margin bounds explain why weighted votes generalize after training error vanishes, but dependent observations require concentration arguments that keep the dependence model and failure budget explicit. We study stationar...