Re: Linear approx for cube root?

Toby Kelsey wrote:
For a minimax solution, the error has maximum absolute value at the lower and
upper ends, and somewhere in the middle, with alternating sign.

I see that the Chebyshev/Borel theorem says that the error takes on its
maximum value 3 times, with alternating signs. How do we know that the
lower and upper ends are two of these points (x_hi and x_lo in your
notation)?

.