So I've been given a piece of pseudo code that involves nested loops. The answer to this question is Θ(n^5) but I do not understand why it is so.

What is the time complexity of this code?

for  i = 1 to n   
   for  j = 1 to n*n    
      for  k = 1 to j     

where the time complexity of Task is Θ(1)

The most inner loop, for k = 1 to j is Θ(j). Then the middle loop, for j = 1 to n^2 I thought, should produce Θ(n^2) as it loops from j = 1 to n^2 but I've been told it's Θ(n^4). I honestly do not understand how it could be Θ(n^4) when it's iterating from 1 to n^2. Can someone explain how this works in simple terms?

  • 5
    $\begingroup$ See here for a structured approach to algorithm analysis. There are also many questions about algorithm-analysis+loops. $\endgroup$
    – Raphael
    Sep 20, 2016 at 7:46

3 Answers 3


You are correct: the middle loop executes $\Theta(n^2)$ inner loops.

What you are told is correct: the runtime of the middle loop is $\Theta(n^4)$.

Your mistake is confusing the two. The inner loop does not run in constant time, so you should not expect the middle loop to complete in $\Theta(n^2)$ time. You can, however, expect the middle loop to complete in

$$ \sum_{j=1}^{n^2} \Theta(j) = \Theta\left( \sum_{j=1}^{n^2} j \right) = \Theta(n^4)$$

time. (and with care, you can even make this calculation rigorous!)

You could heuristically guess that enough loop iterations take close enough to maximum time and guess the middle loop would take $n^2 \cdot \Theta(n^2)$ time, although this heuristic can go astray.

However, the heuristic can also be made rigorous -- e.g. by determining $\Theta(n^2)$ of the loop iterations will complete in half of the maximum time, or by determining that the last half of the loop iterations each take $\Theta(n^2)$ time.


Translating for-loops into sums is the best way to approach this as the resulting calculations are elementary. We get:

$\qquad\displaystyle\begin{align*} \#\mathrm{Task} \quad&= &&\sum_{i=1}^n \sum_{j=1}^{n^2} \sum_{k=1}^j 1 \\ &= && \sum_{i=1}^n \sum_{j=1}^{n^2} j \\ &= && \sum_{i=1}^n \frac{n^2(n^2 + 1)}{2} \\ &= && \frac{n^3(n^2 + 1)}{2}. \end{align*}$

We use the closed form of the triangular numbers (which you find on any formulary if you have not memorized it). Finally, obtain a $\Theta$-class using standard tools.


For each value of $i$, the number of times that "task" is run is $$ \sum_{j=1}^{n^2} \sum_{k=1}^j 1 = \sum_{j=1}^{n^2} j. $$ The latter sum is at least $$ \sum_{j=1}^{n^2} j \geq \sum_{j=n^2/2}^{n^2} n^2/2 \geq (n^2/2)^2 = \Omega(n^4). $$ It is also plainly at most $(n^2)^2 = O(n^4)$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.