Suppose that we have two functions $f(x)$ and $g(x)$ both are increasing (not constant) on an interval $[a,b]$. We know that $f(x) \ne g(x)$ on the interval, except for measure zero subsets of $[a,b]$. I am wondering if there is any transformation $T$ such that $T(f(x)) = T(g(x))$ almost everywhere on $[a,b]$.

I mean any meaningful transformation, not like $T(\cdot) = c$ for some constant $c$.

  • $\begingroup$ There should be fairly tame pairs of continuous functions where $T$ will be forced to be trivial. I am picturing the graphs of $f$ and $g$ between two points where they agree and then trying to glue/weld values together only as necessary to construct such a $T$. These identifications will look something like orbits, ricocheting back and forth from graph to graph as they sink to the lower common value of $f$ and $g$. A perturbation should allow these trajectories to become too 'tangled' to allow a non-constant function. But you should ask someone who works in one-dimensional real dynamics. $\endgroup$ – John Samples Jul 31 '18 at 9:05
  • $\begingroup$ The question specifies no requirements on $T$, so of course there always exists such a transformation. $\endgroup$ – 6005 Aug 10 '18 at 4:04

You can easily define transformations $T$ such that $T(f(x)) = T(g(x))$ holds for some $f(x) \ne g(x)$. As you point out, a trivial example of $T$ is the $0$ transformation, i.e. $T(f(x)) = 0$ for all functions $f(x)$, then for any two functions $f$ and $g$, $T(f(x)) = T(g(x))$. For a less trivial example, let $f(x) = x$, $g(x) = x + 1$, and let $T$ be differentiation. Then $f(x) \ne g(x)$ almost everywhere, but $T(f(x)) = T(g(x)) = 1$ for all $x$.

If you mean for $T$ to be a linear transformation, then what you are interested in is the nullspace of $T$. In particular, if $f(x) - g(x)$ is in the nullspace of $T$, then we will have $T(f(x) - g(x)) = 0$, which implies $T(f(x)) = T(g(x))$. It is fairly common for linear transformations to have a non-trivial nullspace, for example differentiation, like I noted above.

If you are not just looking at linear transformations, I don't know if there's a specific terminology for cases when $T(f(x)) = T(g(x))$, but you're essentially asking about the existence of non-injective transformations (since injectivity is equivalent to $f(x) \ne g(x) \implies T(f(x)) \ne T(g(x))$), but there is no reason to expect an arbitrary transformation to be injective.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.