0
$\begingroup$

I'm found this claim in something I was reading and I'm having trouble working out the logic. Can someone help me out?

$\endgroup$
  • 1
    $\begingroup$ Can you confirm the $a \log m$ part? Is that $a \log a$, perhaps? $\endgroup$ – chi Jan 1 '17 at 13:15
1
$\begingroup$

$m \leq a \log m + b$ has no asymptotic behavior. This is because regardless of constants $a, b$, the left hand side grows faster than the right side, leaving only a finite range where $m$ is valid.

It is likely what was meant is $m \leq a \log a + b$. In this case we have that $m$ is bounded by something, so $m = O(a \log a + b)$. If you have that $O(x + y) = O(\max(x, y))$, then the question is solved.

But why is $O(x + y) = O(\max(x, y))$ true? With big-$O$ notation we are interested in the asymptotically fastest growing part. We ignore constant factors. Now check three cases (in the limit):

$$x < y \rightarrow O(x + y) = O(y)$$ $$x > y \rightarrow O(x + y) = O(x)$$ $$x = y \rightarrow O(x + y) = O(x + x) = O(2x) = O(x)$$

Since our choice for $x = y$ doesn't matter, we can see that choosing the minimum gives us the right asymptotic behavior.

$\endgroup$
  • $\begingroup$ Since the result is $O(\max(a\log a, b))$ is guess that $a,b$ are not constants. It's not entirely clear, though. $\endgroup$ – chi Jan 1 '17 at 19:17

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.