2

Let $A$ and $B$ be nonempty, closed and convex subsets of a Hilbert space $H$. Let $\alpha, \beta \in (0,1)$ such that $\alpha + \beta <1 $. Define $T:H \rightarrow H$ by $$ Tx = \alpha P_A x + \beta P_B x .$$ Show that $T$ is a contraction.

Here, $P_A$ and $P_B$ are the projection operators onto $A$ and $B$, respectively.

I need to show that there exists $\gamma \in [0,1)$ such that $$ d(Tx,Ty) \leq \gamma d(x,y) \text{ for every $x,y\in H$}. $$

Let $x,y \in H$. Then $$d(Tx,Ty) = d(\alpha P_A x + \beta P_B x, \alpha P_A y + \beta P_B y)$$

Please help. Thank you.

Lea
  • 49

1 Answers1

2

since you are in a Hilbert space, you can you the norm : $$||\alpha P_Ax+\beta P_Bx-\alpha P_Ay-\beta P_By||$$

$$\le \alpha|| P_Ax-P_Ay||+\beta|| P_Bx-P_By||$$ Since $P_A$ and $P_B$ are projection on convex closed set in a Hilbert space, they are contractant (see this answer for a demonstration)

Finally : $$||\alpha P_Ax+\beta P_Bx-\alpha P_Ay-\beta P_By||\le (\alpha+\beta)|| x-y||$$

stity
  • 3,500
  • The projection operator does not have to be linear. Also, it does not have to hold $|Px| \le |x|$. Consider the projection on the closed subset ${x_0} \subseteq H$. It is the constant function $x_0$. You are probably thinking of projections onto linear subspaces, but then the proof is easy: $$|T| = |\alpha P_A + \beta P_B| \le \alpha |P_A| + \beta |P_B| = \alpha + \beta$$ so for every $x, y \in H$ we have $$|Tx - Ty| = |T(x-y)| \le |T||x-y| \le \underbrace{(\alpha + \beta)}_{<1}|x-y|$$ – mechanodroid Dec 11 '17 at 16:33
  • @mechanodroid I was indeed thinking of linear spaces but I edited my answer – stity Dec 11 '17 at 17:59