Agnostic Learning of Halfspaces with Gradient Descent via Soft Margins
Spencer Frei 1 Yuan Cao 2 Quanquan Gu 2
Abstract minimize the surrogate risk
We analyze the properties of gradient descent on F` (w) := E(x,y)D `(yw> x) .
(1)
convex surrogates for the zero-one loss for the ag-
nostic learning of halfsp ...


雷达卡




京公网安备 11010802022788号







