Big-O for Eight Year Olds? [duplicate]

I'm asking more about what this means to my code. I understand the concepts mathematically, I just have a hard time wrapping my head around what they mean conceptually. For example, if one were to perform an O(1) operation on a data structure, I understand that the number of operations it has to perform won't grow because there are more items. And an O(n) operation would mean that you would perform a set of operations on each element. Could somebody fill in the blanks here?

• Like what exactly would an O(n^2) operation do?
• And what the heck does it mean if an operation is O(n log(n))?
• And does somebody have to smoke crack to write an O(x!)?

One way of thinking about it is this:

O(N^2) means for every element, you're doing something with every other element, such as comparing them. Bubble sort is an example of this.

O(N log N) means for every element, you're doing something that only needs to look at log N of the elements. This is usually because you know something about the elements that let you make an efficient choice. Most efficient sorts are an example of this, such as merge sort.

O(N!) means to do something for all possible permutations of the N elements. Traveling salesman is an example of this, where there are N! ways to visit the nodes, and the brute force solution is to look at the total cost of every possible permutation to find the optimal one.

The big thing that Big-O notation means to your code is how it will scale when you double the amount of "things" it operates on. Here's a concrete example:

Big-O       |  computations for 10 things |  computations for 100 things
----------------------------------------------------------------------
O(1)        |   1                         |     1
O(log(n))   |   3                         |     7
O(n)        |  10                         |   100
O(n log(n)) |  30                         |   700
O(n^2)      | 100                         | 10000

So take quicksort which is O(n log(n)) vs bubble sort which is O(n^2). When sorting 10 things, quicksort is 3 times faster than bubble sort. But when sorting 100 things, it's 14 times faster! Clearly picking the fastest algorithm is important then. When you get to databases with million rows, it can mean the difference between your query executing in 0.2 seconds, versus taking hours.

Another thing to consider is that a bad algorithm is one thing that Moore's law cannot help. For example, if you've got some scientific calculation that's O(n^3) and it can compute 100 things a day, doubling the processor speed only gets you 125 things in a day. However, knock that calculation to O(n^2) and you're doing 1000 things a day.

clarification: Actually, Big-O says nothing about comparative performance of different algorithms at the same specific size point, but rather about comparative performance of the same algorithm at different size points:

                 computations     computations       computations
Big-O       |   for 10 things |  for 100 things |  for 1000 things
----------------------------------------------------------------------
O(1)        |        1        |        1        |         1
O(log(n))   |        1        |        3        |         7
O(n)        |        1        |       10        |       100
O(n log(n)) |        1        |       33        |       664
O(n^2)      |        1        |      100        |     10000

You might find it useful to visualize it:

Also, on LogY/LogX scale the functions n^1/2, n, n² all look like straight lines, while on LogY/X scale 2ⁿ, eⁿ, 10ⁿ are straight lines and n! is linearithmic (looks like n log n).

This might be too mathematical, but here's my try. (I am a mathematician.)

If something is O(f(n)), then it's running time on n elements will be equal to A f(n) + B (measured in, say, clock cycles or CPU operations). It's key to understanding that you also have these constants A and B, which arise from the specific implementation. B represents essentially the "constant overhead" of your operation, for example some preprocessing that you do that doesn't depend on the size of the collection. A represents the speed of your actual item-processing algorithm.

The key, though, is that you use big O notation to figure out how well something will scale. So those constants won't really matter: if you're trying to figure out how to scale from 10 to 10000 items, who cares about the constant overhead B? Similarly, other concerns (see below) will certainly outweigh the weight of the multiplicative constant A.

So the real deal is f(n). If f grows not at all with n, e.g. f(n) = 1, then you'll scale fantastically---your running time will always just be A + B. If f grows linearly with n, i.e. f(n) = n, your running time will scale pretty much as best as can be expected---if your users are waiting 10 ns for 10 elements, they'll wait 10000 ns for 10000 elements (ignoring the additive constant). But if it grows faster, like n², then you're in trouble; things will start slowing down way too much when you get larger collections. f(n) = n log(n) is a good compromise, usually: your operation can't be so simple as to give linear scaling, but you've managed to cut things down such that it'll scale much better than f(n) = n².

Practically, here are some good examples:

• O(1): retrieving an element from an array. We know exactly where it is in memory, so we just go get it. It doesn't matter if the collection has 10 items or 10000; it's still at index (say) 3, so we just jump to location 3 in memory.
• O(n): retrieving an element from a linked list. Here, A = 0.5, because on average you''ll have to go through 1/2 of the linked list before you find the element you're looking for.
• O(n²): various "dumb" sorting algorithms. Because generally their strategy involves, for each element (n), you look at all the other elements (so times another n, giving n²), then position yourself in the right place.
• O(n log(n)): various "smart" sorting algorithms. It turns out that you only need to look at, say, 10 elements in a 10¹⁰-element collection to intelligently sort yourself relative to everyone else in the collection. Because everyone else is also going to look at 10 elements, and the emergent behavior is orchestrated just right so that this is enough to produce a sorted list.
• O(n!): an algorithm that "tries everything," since there are (proportional to) n! possible combinations of n elements that might solve a given problem. So it just loops through all such combinations, tries them, then stops whenever it succeeds.

don.neufeld's answer is very good, but I'd probably explain it in two parts: first, there's a rough hierarchy of O()'s that most algorithms fall into. Then, you can look at each of those to come up with sketches of what typical algorithms of that time complexity do.

For practical purposes, the only O()'s that ever seem to matter are:

• O(1) "constant time" - the time required is independent of the size of the input. As a rough category, I would include algorithms such as hash lookups and Union-Find here, even though neither of those are actually O(1).
• O(log(n)) "logarithmic" - it gets slower as you get larger inputs, but once your input gets fairly large, it won't change enough to worry about. If your runtime is ok with reasonably-sized data, you can swamp it with as much additional data as you want and it'll still be ok.
• O(n) "linear" - the more input, the longer it takes, in an even tradeoff. Three times the input size will take roughly three times as long.
• O(n log(n)) "better than quadratic" - increasing the input size hurts, but it's still manageable. The algorithm is probably decent, it's just that the underlying problem is more difficult (decisions are less localized with respect to the input data) than those problems that can be solved in linear time. If your input sizes are getting up there, don't assume that you could necessarily handle twice the size without changing your architecture around (eg by moving things to overnight batch computations, or not doing things per-frame). It's ok if the input size increases a little bit, though; just watch out for multiples.
• O(n^2) "quadratic" - it's really only going to work up to a certain size of your input, so pay attention to how big it could get. Also, your algorithm may suck -- think hard to see if there's an O(n log(n)) algorithm that would give you what you need. Once you're here, feel very grateful for the amazing hardware we've been gifted with. Not long ago, what you are trying to do would have been impossible for all practical purposes.
• O(n^3) "cubic" - not qualitatively all that different from O(n^2). The same comments apply, only more so. There's a decent chance that a more clever algorithm could shave this time down to something smaller, eg O(n^2 log(n)) or O(n^2.8...), but then again, there's a good chance that it won't be worth the trouble. (You're already limited in your practical input size, so the constant factors that may be required for the more clever algorithms will probably swamp their advantages for practical cases. Also, thinking is slow; letting the computer chew on it may save you time overall.)
• O(2^n) "exponential" - the problem is either fundamentally computationally hard or you're being an idiot. These problems have a recognizable flavor to them. Your input sizes are capped at a fairly specific hard limit. You'll know quickly whether you fit into that limit.

And that's it. There are many other possibilities that fit between these (or are greater than O(2^n)), but they don't often happen in practice and they're not qualitatively much different from one of these. Cubic algorithms are already a bit of a stretch; I only included them because I've run into them often enough to be worth mentioning (eg matrix multiplication).

What's actually happening for these classes of algorithms? Well, I think you had a good start, although there are many examples that wouldn't fit these characterizations. But for the above, I'd say it usually goes something like:

• O(1) - you're only looking at most at a fixed-size chunk of your input data, and possibly none of it. Example: the maximum of a sorted list.
  • Or your input size is bounded. Example: addition of two numbers. (Note that addition of N numbers is linear time.)
• O(log n) - each element of your input tells you enough to ignore a large fraction of the rest of the input. Example: when you look at an array element in binary search, its value tells you that you can ignore "half" of your array without looking at any of it. Or similarly, the element you look at gives you enough of a summary of a fraction of the remaining input that you won't need to look at it.
  • There's nothing special about halves, though -- if you can only ignore 10% of your input at each step, it's still logarithmic.
• O(n) - you do some fixed amount of work per input element. (But see below.)
• O(n log(n)) - there are a few variants.
  • You can divide the input into two piles (in no more than linear time), solve the problem independently on each pile, and then combine the two piles to form the final solution. The independence of the two piles is key. Example: classic recursive mergesort.
  • Each linear-time pass over the data gets you halfway to your solution. Example: quicksort if you think in terms of the maximum distance of each element to its final sorted position at each partitioning step (and yes, I know that it's actually O(n^2) because of degenerate pivot choices. But practically speaking, it falls into my O(n log(n)) category.)
• O(n^2) - you have to look at every pair of input elements.
  • Or you don't, but you think you do, and you're using the wrong algorithm.
• O(n^3) - um... I don't have a snappy characterization of these. It's probably one of:
  • You're multiplying matrices
  • You're looking at every pair of inputs but the operation you do requires looking at all of the inputs again
  • the entire graph structure of your input is relevant
• O(2^n) - you need to consider every possible subset of your inputs.

None of these are rigorous. Especially not linear time algorithms (O(n)): I could come up with a number of examples where you have to look at all of the inputs, then half of them, then half of those, etc. Or the other way around -- you fold together pairs of inputs, then recurse on the output. These don't fit the description above, since you're not looking at each input once, but it still comes out in linear time. Still, 99.2% of the time, linear time means looking at each input once.