# algorithm complexity calculation T(n) = 2T(n/2) + n*log(n)

I guess I lack at understanding basic things at algorithm calculations, while learning for an exam.

• at the result, does one write like O(ld n) or just log instead of ld ?

• regarding the following calculation (not from me) at the yellow mark - why is it Theta(n) ? why not O(n) ? I dont get it.. big thanks in advance !

## 1 Answer

Let's start at the beginning.

Basically, there are 3 very popular notations to express time complexity of algorithms:

• $\Theta(g(n))$,
• $\mathcal{O}(g(n))$ (this is the well-known Big O notation),
• $\Omega(g(n))$.

The first thing that is in most cases a little bit confusing (and misused), that these notations denote sets - sets of functions. For example, the interpretation of $\Theta(g(n))$ is as follows:

$$\Theta(g(n)) = \{ f(n) \; | \; \text{There exist } n_0, c_1 \text{ and } c_2 \text{ constants so that } 0 \leq c_1 \cdot g(n) \leq f(n) \leq c_2 \cdot g(n) \text{ for all } n > n_0. \}$$

So, $\Theta(g(n))$ is a set of $f(n)$ functions for that $g(n)$ can be used as both an upper and a lower bound (with the use of the given $c_1$ and $c_2$ constants). In other words, if you plot these 3 functions, $f(n)$ is going to be between $c_1 \cdot g(n)$ and $c_2 \cdot g(n)$ - at least, for all inputs larger than or equal to $n_0$. (It's worth taking a look at figures about these in Google so that you can get a better understanding of it.)

Note: because of the use of $n_0$ and all larger inputs than it, these notations are also called asymptotic notations.

The interpretation of $\mathcal{O}(g(n))$ and $\Omega(g(n))$ are very similar to the one above, but they only refer to either the upper bound or the lower bound, respectively. (Of course, the existence of only one $c$ constant is enough for these 2 definitions.)

Just in case, see the definition of $\mathcal{O}(g(n))$ below:

$$\mathcal{O}(g(n)) = \{ f(n) \; | \; \text{There exist } n_0 \text{ and } c \text{ constants so that } 0 \leq f(n) \leq c \cdot g(n) \text{ for all } n > n_0. \}$$

In other words, $g(n)$ is an upper bound for all the functions in the set $\mathcal{O}(g(n))$. For example, let's denote the time complexity of the insertion sort with $T(n) = \frac{n(n - 1)}{2} = \frac{n^2}{2} - \frac{n}{2}$ (you can derive this easily if you think through the algorithm). Now, here $T(n) \in \mathcal{O}(n^2)$ means that the time complexity of the algorithm is quadratic - so if the size of the input is $n$ (we have an array to be sorted consisting of $n$ elements), the algorithm must perform the most expensive operation at most $c \cdot n^2$ times (where $c$ might be 1). Three notes here:

• Typically, we don't care about the "weaker" members of the equations, like $\frac{n}{2}$ in the example above, because of the asymptotic property I mentioned earlier (that wants to say something like "for large enough inputs always the strongest members will only count").
• Typically, we only consider the most expensive operations of an algorithm, i.e., in the case of sorting algorithms, the number of comparisons.
• Many times, people use the notations of $T(n) \in \mathcal{O}(g(n))$ and $T(n) = \mathcal{O}(g(n))$ like they were equivalent (basically, they aren't, but it's a common thing since the use of the latter can be advantageous as well).

So now, that hopefully we are done with the basics, you can see, that if a function $f(n) \in \Theta(g(n))$, it is also true that: $f(n) \in \mathcal{O}(g(n))$ and $f(n) \in \Omega(g(n))$ (in most textbooks, this is presented as a theorem with its proof, as well). In most cases, people don't care about the lower bounds of an algorithm's time complexity - so, it shouldn't really matter whether you see $\mathcal{O}$ or $\Theta$ (this would be the answer to your second question). I think that $\Omega$ is much less frequent than the other two.

I haven't ever seen the $ld$ notation you mentioned. However, I would prefer to use $\log_2$, it will be clear for anyone. Of course, you can use the one you would like to, but make sure to put them into either $\mathcal{O}$ or $\Theta$ (I can't recall a case when only the function was indicated).

Truly hope that I managed to provide a detailed answer you find useful enough. If you have any further question, please feel free to ask.

• I find this answer misleading (the question is not about algorithms) and overly verbose: let's assume, for instance, that the OP has access to a definition of Landau notation. I know you mean well, but reproducing textbook chapters on every other question is not a good use of (y)our time. – Raphael Jun 9 '17 at 20:05
• You probably right, thank you for the advices. Despite I wrote quite a long answer that contains examples for algorithms, I still tried to focus on the notations. Sorry, if it became misleading, will do my best next time. – laszlzso Jun 9 '17 at 20:32
• How does this answer the question? – David Richerby Jun 9 '17 at 20:33
• One of the questions was whether there should be $\Theta$ or $\mathcal{O}$. I could have written only that it doesn't really matter but I thought it a good idea to give a detailed explanation. – laszlzso Jun 9 '17 at 20:37
• @ZsoltLászló thanks for the detailed answer, i really appreciate that !! imagine this: sum (x^i) * O(1), from i=0 to k-1 x can be anything, i'm more interested in the O(1). if k is equal to n, for example, the result of this part would be O(n). But in this case i would then write Theta(n) instead O(n), right ? I guess so, because it will be a harder limit.. – Shoorty Jun 11 '17 at 6:00