Binary classification. Consider a binary classification problem with data in pairs $($xi$,$ yi$)$ in Rn $\backslash$times $\{1,1\},$

where we represent a classifer mapping x in Rn to $\{\backslash$pm $1\}$ by $\backslash$theta in Rn$,$ predicting $$y $=$ sign$($xT $\backslash$theta $).$ Given m

observations $($xi$,$ yi$),$ i $=1,...,$ m$,$ we solve the convex optimization problem