- #1
diegzumillo
- 173
- 18
- TL;DR Summary
- Average of number of prime factors have a curious fit, and I don't know what it means.
Hi there, bored physicist here. I wondered about factorization so I made some plots and a model fit to satiate my curiosity, but it only made me more curious. Now I just want the answers! I didn't know which prefix to use. "Hobbyist" seemed more appropriate but we don't have that so I went with "I" (undergrad level). Hope that works.
So I started calculating how many prime factors a number has, then plotted this number for all the numbers up to some arbitrary high number (I went up to a million). The plot was not very informative. (this one has 10000, for higher numbers the plot becomes an unreadable series of lines)
This looks like a random function. But, obviously we will start seeing higher number of factors the higher the number, so maybe if I take the average of the number of factors every batch of numbers that will tell me something interesting. So I experimented with different batch sizes (steps) to see if that influenced the result much, and it doesn't. It just narrows the curve a little, which is nice.
Bigger step/batch:
Naturally I wanted to know what curve this is, so I tried some naive fits. It looks like a square root of x so I tried a x^b. Close but definitely not it. Tried log, also close but not it. Then I went to wikipedia and, even though most of those number theory articles flew over my head, I did notice a recurrence of log(log(x)), so I tried that and got 1.27568 log ( log (815024. x - 600100.))
All right! that looks spot on!
but why? What are these numbers? Changing the step makes no difference on this fit, by the way.
So I started calculating how many prime factors a number has, then plotted this number for all the numbers up to some arbitrary high number (I went up to a million). The plot was not very informative. (this one has 10000, for higher numbers the plot becomes an unreadable series of lines)
This looks like a random function. But, obviously we will start seeing higher number of factors the higher the number, so maybe if I take the average of the number of factors every batch of numbers that will tell me something interesting. So I experimented with different batch sizes (steps) to see if that influenced the result much, and it doesn't. It just narrows the curve a little, which is nice.
Bigger step/batch:
Naturally I wanted to know what curve this is, so I tried some naive fits. It looks like a square root of x so I tried a x^b. Close but definitely not it. Tried log, also close but not it. Then I went to wikipedia and, even though most of those number theory articles flew over my head, I did notice a recurrence of log(log(x)), so I tried that and got 1.27568 log ( log (815024. x - 600100.))
All right! that looks spot on!
but why? What are these numbers? Changing the step makes no difference on this fit, by the way.