Consider the inequality

$$ c_1\lg(n) \leq \lg(k) \leq c_2\lg(n),\text{ for } n \geq n_{0} $$

With $c_1,c_2,n_0 > 0$, $\lg(k) = \Theta(\lg(n))$

By deriving the actual relationship of $k$ with $n$, we have

$$ \Rightarrow 2^{c_1\lg(n)} \leq 2^{\lg(k)} \leq 2^{c_2\lg(n)}\\ \Rightarrow 2^{\lg(n^{c_1})} \leq 2^{\lg(k)} \leq 2^{\lg(n^{c_2})}\\ \Rightarrow n^{c_1} \leq k \leq n^{c_2} $$

If $c_1 \neq c_2$, is it still possible to provide tight bounds (e.g. Theta notation) for $k$ given that $k = \Omega(n^{c_1})$ and $k = O(n^{c_2})$?

  • 1
    $\begingroup$ No, nothing more precise than $\Omega(n^{c_1})$ and $O(n^{c_2})$. $\endgroup$ – Did Mar 1 '12 at 7:20

Consider $n$, $n^2$ and $n^3$.

We have that $\log n^2 = \theta(\log n)$ but, $n^2 \neq \theta(n^3)$ and $n^2 \neq \theta(n)$.

Does that answer your question?

  • $\begingroup$ yes, I think cases where $lg(k) = \Theta(lg(n)) \Rightarrow k = \Theta(n)$ are only rare and must not be used for generalization. $\endgroup$ – edgar Mar 1 '12 at 7:25

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.