1: \begin{proof}

2: The estimate is illustrated in the Figure \ref{fig:L1norm}. Here, we give a rigorous proof.

3: \\

4: Set

5: \begin{align*}

6: P_k:=\{x_i\in \Nhi:\ k\sigma \le |s_h(x_i)|<(k+1)\sigma\}.

7: \end{align*}

8: %

9: Then we clearly have

10: \begin{align*}

11: \sum_{x_i\in \Nhi} f_i |s_h(x_i)|  

12: = \sum_{k=0}^N \sum_{x_i\in P_k} f_i |s_h(x_i)|

13: \le \sum_{k=0}^N (k+1)\sigma \mu(P_k).

14: \end{align*}

15: %

16: We also have

17: \begin{align*}

18: A_k = \bigcup_{m\ge k}^N P_m,

19: \end{align*}

20: and so, since the sets $\{P_k\}$ are disjoint,

21: \begin{align*}

22: \mu(A_k) = \sum_{m=k}^N \mu(P_m).

23: \end{align*}

24: Therefore

25: \begin{align*}

26: \sigma \sum_{k=0}^N \mu(A_k) = \sigma \sum_{k=0}^N \sum_{m=k}^N \mu(P_m) = \sigma \sum_{k=0}^N (k+1) \mu(P_k) \ge  \sum_{x_i\in \Nhi} f_i |s_h(x_i)|.

27: \end{align*}

28: \hfill

29: \end{proof}

30: