1: \begin{proof}

2: Without loss of generality, we will assume $f(0) = 0$.

3: Let $\tau = \TV(f')$.

4: As $f'$ has bounded total variation, there exists a measure $\mu$ on $[0,1]$ such that $f(x) = f_\mu$: \[f(x) =  \int{(x - t)}_+ \;d \mu(t).\]

5: 

6: That is, $f$ is a spline with infinitely many knots. The idea is to use Caratheodory's theorem for convex hulls to see that, as we only care about $\mu$ in terms of its action on a finite number of functions (basically, we only care about the values of $f$ at $x_i$), we can replace $\mu$ with a measure supported on finitely many points.

7: 

8: To make this idea rigorous, note that the vector

9: \[v = (f(x_1), \ldots, f(x_n), 0) = \int(  (x_1 - t)_+, \ldots, (x_n - t)_+, 1) \;d \mu(t) \]

10: must lie in convex hull of the (convex) set

11: \[ C = \{\pm( \tau (x_1 - t)_+, \ldots,  \tau (x_n - t)_+,  \tau) : t \in [0,1] \} \subset \reals^{n+1}\]

12: as $\| \mu \|_1 = \tau$.

13: Caratheodory's theorem for convex hulls ensures us that $v$ can be represented as a convex combination of at most $n+2$ points from $C$. Letting these $n+2$ points be represented by their indicies, $t_1, \ldots, t_{n+2}$, and their weights $\alpha_1, \ldots , \alpha_{n+2}$ we define $w_j = \alpha_j \tau$ to obtain:

14: \begin{align*}

15: f(x_i)  &= \sum_j w_j (x_i - t_j)_+ = f_\mu(x_i) \\

16: \sum_j w_j &= 0.

17: \end{align*}

18: Here $\mu = \sum_j w_j \delta_{t_j}$.

19: As $\sum_j | w_j | = \tau$, we have $\TV(f_\mu') = \| \mu \|_1 = \tau$.

20: 

21: 

22: \end{proof}

23: