Lychrel numbers

Blogging burns a lot of caffeine. If you enjoy my posts, you should buy me a cup of tea.

Last night I discovered another cool mathematical concept akin to the Collatz conjecture - Lychrel numbers.

Collatz map fractal in a neighbourhood of the ...

Image via Wikipedia

The idea of a lychrel number is pretty straightforward: Take a number, add its reverse, continue until you reach a palindrome number. If you never reach a palindrome, then this is a Lychrel number.

Something like this:

349 + 943 = 1292,
1292 + 2921 = 4213
4213 + 3124 = 7337
not lychrel

If you’ve ever done any theoretical computer science, you’ll immediately spot a problem. This isn’t a very good algorithm. Problem is with that “never” word in the description – an algorithm is a finite set of steps, when you need an infinite amount of steps to reach a conclusion that’s … not very useful.

Honestly I am not certain what class of problems lychrel numbers fall into. The “not a lychrel number” is a half-decidable problem. It will always tell you when a number is not lychrel but it will never terminate when it is. If my understanding is correct, this would make “is a lychrel number” an non-decidable problem.

Project Euler is kind enough to limit the problem a little bit and make it a fun algorithm to write before bed when your brain is half dead. Find all lychrel candidates under 10,000 assuming it should never take more than 50 iterations.

Solving that problem becomes rather trivial in Haskell:

 
reverse' = read . reverse . show
 
palindrome n = n == reverse' n
 
-- max denotes max recursion depth
lychrel n max
  | max <= 0 = True
  | palindrome$n+r = False
  | otherwise = lychrel (n+r) (max-1)
    where r = reverse' n
 
lychrels max =
  length [x | x <- [1..max], lychrel x 50]

Oh and actually the first number that needs more than 50 iterations to converge into a palindrome is 10677, so the problem is pretty safely stated.

For a final bit of fun, the number 4994, itself a palindrome, is a lychrel candidate.

Enhanced by Zemanta

---
Need a freelance developer? Email me!

You should follow me on twitter
 Subscribe to RSS

3 responses so far

  • http://twitter.com/zidarsk8 Miha Zidar

    These numbers are fun. But as far as I know, no one has yet to prove the fact that this is an undecidable problem. Maybe it’s just that the algorithms we have, really suck :P

    Anyway the haskel solution sure looks nicer than my java code.

  • anon

    For the problem of determining if a number is lychrel to be undecidable would mean that no Turing machine (or equivalently no algorithm) exists that can decide it.

    You’re providing an algorithm that doesn’t decide the problem (it’s only an enumerator, not a decider). That doesn’t mean that there is no deciding algorithm, though.

  • http://swizec.com Swizec

    Sure hope someone proves their existence soon, then maybe we can start working on that turing machine to find them :)

« A month wasted Heroku, mongo, node.js - a problem »