7
$\begingroup$

Diagonalization is a frequently used technique in complexity theory. However, the problems (sets) that are created by diagonalization rarely correspond to anything natural. It would be interesting to collect examples, when a problem was originally defined via diagonalization, but later it was found to coincide with a natural problem.

$\endgroup$
6
  • $\begingroup$ it is in a way hard to argue that any TM complete problem is any more "natural" than any other, they are equivalent and there seems to be no "natural" way of discriminating them except perhaps by # of states in the TMs etc.... $\endgroup$
    – vzn
    Feb 8, 2014 at 1:45
  • $\begingroup$ fairly close to easy to understand undecidable problem, Mathematics $\endgroup$
    – vzn
    Feb 8, 2014 at 1:56
  • $\begingroup$ similar also to simplest TM complete system(s) tcs.se $\endgroup$
    – vzn
    Feb 8, 2014 at 2:05
  • $\begingroup$ @vzn: The question didn't say anything about completeness... $\endgroup$ Feb 8, 2014 at 2:17
  • $\begingroup$ ok but it did not rule it out either as in your answer... $\endgroup$
    – vzn
    Feb 8, 2014 at 2:19

1 Answer 1

7
$\begingroup$

In some sense the halting problem satisfies this request (I was almost ready to say "There are no such examples!" when I thought of this one.) By doing diagonalization against computable sets, you can of course get sets of very high complexity, but the seemingly-most-natural way of diagonalizing against computable sets gives you a problem that is equivalent to the halting problem. As we now know, the halting problem is equivalent to many other computational problems, such as deciding many properties of TMs, telling whether a Diophantine equation has solutions, the word problem in finitely presented groups, telling whether a high-dimensional knot is the unknot, etc.

I am still ready to say that I would be surprised to see other examples (that aren't just, say, relativized versions of this).

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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