Jumpstart - Questions and Answers Week 3
This week's discussion
takes place on Overleaf.
Question 1 (James Shade, incoming Fall 2018)
I am having a hard time coming up with a good intuitive understanding
of what uniform continuity represents. I understand the definition and
how it differs from the definition of standard continuity, but I am
having a hard time identifying is the bigger picture of the
information uniform continuity captures. I do notice a relationship
between uniform continuity of a differentiable function and its
derivative. My thinking is this: if the function's derivative is
unbounded, then for any $\delta$, we should be able to choose $x$ and
$y$ such that $d(x,y) < \delta$, but $d(f(x),f(y))$ as far apart as we
desire. Hence, the function cannot be uniformly continuous. Is my
thinking on this correct? Can anyone help me determine the right way
to conceptualize uniform continuity?
Answer
Followup question (Matthew Hirning, incoming Fall 2018)
I was also thinking about uniform continuity in terms of a bounded
derivative, and the nexample of $x^2$ makes perfect sense. I was
wondering if anyone could explain conceptually why the unbounded
derivative in the case of $\sqrt{x}$ is not an issue. I'm trying to
envision it and am having trouble.
What you are saying is actually true if the function were linear: if
the slope becomes infinite, the function looses continuity. For
general functions, though, the slope of the tangent at a point can get
large without the function loosing continuity. Second what happens at
a point can happen anywhere. There are pretty bad functions, like,
continuous everywhere but differentiable nowhere (think of a path of
Brownian motion, which is Hölder continuous but nowhere
differentiable).
As for uniform continuity, think of the fact that continuous functions
are uniformly continuous on compact sets. Thus continuous, but not
uniformly so, functions have to be looked for on non-compact sets. In finite
dimensions, this means on non-closed or non-bounded sets.
Comment (Michael Hehman, incoming Fall 2018)
I'm not sure that it is true that if a function has unbounded
derivative, then it is not uniformly continuous. For example, $f(x) =
\sqrt{x}$ is uniformly continuous on $(0, \infty)$, despite the fact
that $f'(x)$ is unbounded on $(0,\infty)$. There is a relationship
between uniform continuity and the derivative being bounded, but it
goes the other way: if a function has bounded derivative on an
interval, then it is uniformly continuous on that set. I would also be
curious as to how others conceptualize uniform continuity though. I
also tend to think about it in terms of derivatives. For example,
$x^2$ is not uniformly continuous since as $|x|$ gets large, $x^2$
becomes arbitrarily steep (derivative is unbounded), and so no
$\delta$ can be found that would work for all $x \in \mathbb{R}$ at
once. However, the $\sqrt{x}$ example shows that unbounded derivative
is not always sufficient to determine uniform continuity.
Comment (Wes Whiting, incoming Fall 2018)
Uniform continuity is useful because regular continuity isn't strong
enough to guarantee certain nice conclusions; for example, you might
hope that a continuous function would send Cauchy sequences to Cauchy
sequences, but this is not the case (consider the function
$\frac{1}{x}$ and the sequence $\frac{1}{n}$). A uniformly continuous
function does preserve the Cauchy property . Every function with a
bounded derivative is Lipschitz by the mean
value theorem, and Lipschitz functions are uniformly
continuous. However, not all uniformly continuous functions are
Lipschitz; Michael gives a good counterexample above. A wider class of
uniformly continuous functions is those which are Hölder-continuous,
but this is still not all of them. Intuitively speaking, a
differentiable function is uniformly continuous as long as the
derivative never gets too large for too long, so that while the
derivative may be very large at a given point, it drops off quickly
enough that you can still bound the change in $y$ for a sufficiently
small change in $x$.
This line of thought, while in principle correct, can
be misleading since it is not precise enough, especially the use of
the expression "too large for too long".
Differentiability is way more than continuity. To make a parallel: if
you look for people who are 5ft or taller, you can also look for
people who are 6ft or taller and you would be fine. You would be
missing very many people, though. Differentiable functions with
bounded derivative are way more special than just continuous: there
are continuous functions that are nowhere differentiable. A finite slope can
give you an indication of how continuous a function is at the point,
but continuity at a point does not imply existence of a tangent
slope. The function
$$
f(x)=\begin{cases} x^\alpha\sin(1/x),&x>0,\\
0,& x=0,\end{cases}
$$
is continuous in $x=0$ for any $\alpha>0$ but check out the derivative
(even for $\alpha=1$).