6
$\begingroup$

I have recently decided to read up on the current integer factorization algorithms. When looking into some of the algorithms, I came across the following statement:

Say that p is the smallest prime factor of n. By Dixon's Theorem, the probability that the largest factor of n is less than $(p)^\epsilon$ is roughly $\epsilon ^ {- \epsilon}$.

But when I hear Dixon's Theorem, I think of this. I currently don't see how Dixon's Theorem yields such a probabilistic bound. Can you help me with that?

And the disclaimer - I am not currently in any class, so this was not assigned to me, it is not a homework problem, and I will in no way be receiving any sort of academic credit for this work.

$\endgroup$
  • $\begingroup$ Congrats on the 10K :-) $\endgroup$ – Asaf Karagila Oct 14 '11 at 8:31
2
$\begingroup$

Wrong Dixon! Please look at this link to get a start. Unfortunately, the relevant original paper (Mathematics of Computation 36, 1981, pp 255-260) is behind a pay wall unless you are at a subscribing university.

$\endgroup$

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.