= Perspective function
{wiki}
For a <function> $f:D\subseteq\mathbb R^n\to\mathbb R$, its perspective is
$$
g(x,t)=t f(x/t),\qquad t>0,\quad x/t\in D.
$$
If $f$ is <convex>, so is $g$ jointly in $(x,t)$. For two domain points and $0<\lambda<1$, put $t=\lambda t_1+(1-\lambda)t_2$ and $\theta=\lambda t_1/t$. Then
$$
\frac{\lambda x_1+(1-\lambda)x_2}{t}=\theta\frac{x_1}{t_1}+(1-\theta)\frac{x_2}{t_2}.
$$
Apply <convexity> of $f$ and multiply by $t$ to obtain the required <convexity> inequality for $g$. The weights are rescaled by the positive $t_i$, not left equal to $\lambda$ and $1-\lambda$.
For $f(x)=\|x\|_2^2$, the perspective is $g(x,t)=\|x\|_2^2/t$, a jointly <convex> quadratic-over-linear <function> for $t>0$. It models a quadratic cost per unit scale. The domain excludes $t=0$; assigning boundary values requires a separate extension.<Convex Optimization by Stephen Boyd and Lieven Vandenberghe>[§3.2.6, printed pp89–90; Example3.18]{ref}
Back to article page