I am trying to prove that if f(n) and g(n) are asymptotically positive functions, then:
f(n) = O((f(n))^2)
f(n) = O(g(n)) implies 2^(f(n)) = O(2^(g(n)))
f(n) = O(g(n)) implies g(n) = O(f(n))
1) Theorem: If f(n) is an asymptotically positive function from natural numbers to natural numbers, then f(n) = O((f(n))^2) (note I have added an extra, perhaps implied, assumption).
Proof: Because f(n) is an asymptotically positive function from natural numbers to natural numbers, it is guaranteed that for all natural numbers n greater than or equal to some natural number n0, f(n) > 0, hence f(n) >= 1. Because f(n) is guaranteed to be positive we are free to multiply both sides of the inequality by f(n) without changing the direction to get f(n)^2 >= f(n). Therefore, we can choose c = 1 and use the n0 from the assumption to show that f(n) = O((f(n))^2). (Recall that by the definition of Big-Oh, f(n) = O(g(n)) if and only if there exist constants c > 0, n0 such that for n >= n0, f(n) <= c * g(n)).
2) Theorem: if f(n) and g(n) are asymptotically positive functions from natural numbers to natural numbers and f(n) = O(g(n)), then it is not necessarily true that 2^(f(n)) = O(2^(g(n)).
Proof: The proof is by example. It can be shown that 4n = O(2n). 4n and 2n are both asymptotically positive functions from naturals to naturals. However, it can also be shown that 2^(4n) = 16^n is not O(2^(2n)) = O(4^n).
3) Theorem: if f(n) and g(n) are asymptotically positive functions from natural numbers to natural numbers and f(n) = O(g(n)), then it is not necessarily true that g(n) = O(f(n)).
Proof: The proof is by example. It can be shown that n = O(n^2). n and n^2 are both asymptotically positive functions from naturals to naturals. However, it can also be shown that n^2 is not O(n).
f(n) = O((f(n))2)
Any function is by default big-O of itself, i.e. we can use a bigger constant cbig such that f(n) <= cbig.f(n).
Thus,
if f(n) is less than or equal to cbig.f(n),
then f(n) will definitely be less than or equal to cbig.f(n).f(n), for asymptotically positive f(n).
Mathematically, f(n) = O(f(n).f(n)) = O(f(n)2) is true.
f(n) = O(g(n)) implies 2f(n) = O(2g(n))
f(n) = O(g(n)) implies that f(n) <= g(n)
Also, if some positive number n is less than m, then 2n will be less than 2m
Using 1. and 2. above, we can conclude that if f(n) = O(g(n)), then 2f(n) = O(2g(n))
f(n) = O(g(n)) implies g(n) = O(f(n))
This one is wrong.
f(n) = O(g(n)) implies g(n) = Ω(f(n)).
If f(n) = O(g(n)), then f(n) is upper bound by g(n) which means that g(n) is lower bound by f(n), therefore g(n) = Ω(f(n)).
Related
I'm fairly certain that if f(n) = Θ(g(n)) is true, if f(n) is asymptotically equal to g(n). However, I'm concerned I might be overlooking something. Am I correct in thinking that f(n) = Θ(g(n)) then f(n) is asymptotically equal to g(n)? or am I overlooking something?
I'm trying to compare different algorithms with respective runtimes of f(n) and g(n) and prove that f(n) = Θ(g(n)), but I'm not sure if I'm on the right way or not.
A. f(n) = log(n^100), g(n) = log(n^2)
lim n->∞ f(n)/g(n) = lim n->∞ log(n^200)/log(n^2) = 100
Since the result is a constant, we conclude that f(n) ∈ ϴ(g(n)), hence f(n) = ϴ(g(n)).
B. f(n) = sqrt(n), g(n) = log(n)
lim n->∞ f(n)/g(n) = lim n->∞ sqrt(n)/log(n) = ±∞, in my case ∞, hence f(n) ≠ ϴ(g(n)).
C. f(n) = 3^n, g(n) = 5^n
lim n->∞ f(n)/g(n) = lim n->∞ 3^n/5^n = 0, hence f(n) ≠ ϴ(g(n)).
D. f(n) = sin(n)+3, g(n) = cos(n)+1
lim n->∞ f(n)/g(n) = lim n->∞ sin(n)+3/cos(n)+1 = 4/3, hence f(n) ≠ ϴ(g(n)).
Please tell me, am I on the right way?
Am I correct in thinking that if f(n) = Θ(g(n)) then f(n) is asymptotically equal to g(n)?
No, asymptotic equality is a stronger claim than asymptotically bound. The opposite is true: when 𝑓(𝑛) is asymptotically equal to 𝑔(𝑛), then 𝑓(𝑛) is Θ(𝑔(𝑛))
As defined on Wikipedia - Asymptotic Analysis:
if and only if
𝑓(𝑛)
lim ──── = 1
𝑥→∞ 𝑔(𝑥)
the functions 𝑓 and 𝑔 are said to be asymptotically equivalent.
For the first example where 𝑓(𝑛) = log𝑛¹⁰⁰ and 𝑔(𝑛) = log𝑛², this equivalence does not hold:
log𝑛¹⁰⁰ / log𝑛² = log𝑛¹⁰⁰ − log𝑛² = log𝑛⁹⁸, whose limit diverges to infinity and so 𝑓(𝑛) and 𝑔(𝑛) are not asymptotically equal.
See also Wikipedia - Family of Bachmann–Landau notations
I'm working through proof of f(n) + o(f(n)) = theta (f(n)) and I came across a part in the proof that I am having trouble understanding.
We let f(n) and g(n) be asymptotically positive functions and assume g(n) = O(f(n)).
In the proof, it states that since we know that f(n) + g(n) ≥ f(n) for all n, we can conclude that f(n) + g(n) = Omega((f(n)).
We can also conclude similarly that f(n) + g(n) ≤ 2 f(n). Therefore f(n) + g(n) = O(f(n)).
I am having trouble understanding why it is the case that f(n) + g(n) = Omega((f(n)) and f(n) + g(n) = O(f(n)) would be true. How is it that we can prove that the tight-lower bound is specifically when we add g(n) to f(n)? What is it that we are exactly concluding from the value of g(n)?
One way of proving that f(n) is theta(g(n)) is to prove two separate statements: that f(n) is omega(g(n)), and f(n) is O(g(n)). It's pretty clear this way of proving is correct from the definitions of those notations.
In this exact problem, if we choose some constant c to be equal to 1, we will have, for every n, that f(n) + g(n) >= c * f(n), so that, by definition, shows that f(n) + g(n) is Omega(f(n)). Furthermore, for the O(f(n)) part, if we choose the constant c to be 2 in this case, we need to prove that there exists some n0 such that f(n) + g(n) <= c * f(n) for every n > n0, which is equivalent to g(n) <= f(n) for every n > n0, which is equivalent to the definition of g(n) = O(f(n)) given in the problem statement.
Hope this helps.
Let's say you were given two logarithmic functions like
and you were asked to find if f(n) is O(g(n)) Ω(g(n)) or Θ(g(n)), how would you go about it? I found questions like these easier when you were comparing two exponential equations, because for example with x(n) = n^2 and p(n) = n^2 you could find a c > 0 (ex 3) where x(n) <= cp(n) for all n greater than some n>0 and that would prove that x(n) = O(p(n)). However, comparing two logarithmic functions seems much more difficult for some reason. Any help is appreciated, thanks!
f(n) is O(g(n)) iff there is a constant c and n_0 such that f(n) <= c * g(n) for each n >= n_0.
f(n) is Ω(g(n)) iff there is a constant c and n_0 such that f(n) >= c * g(n) for each n >= n_0.
Now, f(n) is Θ(g(n)) iff f(n) is O(g(n)) and f(n) is Ω(g(n)).
So, in your cases, we have:
f(n) = log (n^2) = 2logn
which means, g(n) is logn and c = 2, which means f(n) <= 2 * logn and f(n) >= 2 * logn, which makes it Ω(logn).
Btw. its also f(n) <= n and f(n) >= 1, so f(n) can be O(n), but we don't use it, since we can find a better O(g(n)). In this case we don't have the same function in both notations, to for those values we don't have Ω. However, we just need one option for g(n) to declare Ω. In cases we can't find it, we say its not Ω. Note the word "we say".
In second case, we care only for "highest growing value", logn part. Now, c = 1, and g = log(n), so in this case, its also Ω(logn).
I have these three questions for an exam review:
If f(n) = 2n - 3 give two different functions g(n) and h(n) (so g(n) doesn't equal h(n)) such that f(n) = O(g(n)) and f(n) = O(h(n))
Now do the same again with functions g'(n) and h'(n), but this time the function should be of the form
g'(n) = Ɵ(f(n)) and f(n) = o(h'(n))
Is it possible for a function f(n) = O(g(n)) and f(n) = Ω(g(n))?
I know that a function is O(n) of another, if it is less than or equal to the other function. So I think 1. could be g(n) = 2n²-3 and h(n) = 2n²-10.
I also know that a function is Ɵ(n) of another if it is basically equal to the other function (we can ignore constants), and o(n) if it is only less than the function, so for 2. I think you could have g'(n) = 2n-15 and h'(n) = 2n.
To 3.: It is possible for a function to be both O(n) and Ω(n) because O(n) and Ω(n) allows for the function to be the same as the given function, so you could have a function g(n) that equals f(n) and satisfies the rules for being both O and Ω.
Can someone please tell me if this is correct?
Your answers are mostly right. But I would like to add some points:
Given is f(n) = 2n - 3
With g(n) = 2n²-3 and h(n) = 2n²-10 f(n) is in O(g(n)) and in O(h(n)). But your g(n) and h(n) are basicly the same, at least they are both in Θ(n²). There exists many other function that would also work. E.g.
f(n) ∈ O(n) ⇒ g(n) = n
f(n) ∈ O(nk) ⇒ g(n) = nk ∀ k ≥ 1
f(n) ∈ O(2ⁿ) ⇒ g(n) = 2ⁿ
g'(n) = 2n-15 reduces to g'(n) = n, if we think in complexities, and this is right. In fact, it is the only possible answer.
But f(n) ∈ o(h'(n)) does not hold for h'(n) = 2n. Little-o means that
limn → ∞ | f(n)/g(n) | = 0 ⇔ f(n) ∈ o(g(n))
So you can choose h'(n) = n² or more general h'(n) = nk ∀ k > 1 or h'(n) = cⁿ for a constant c > 1.
Yes it is possible and you can take it also as a definition for Θ(g(n)):
f(n) ∈ Θ(g(n)) ⇔ f(n) ∈ O(g(n)) and f(n) ∈ Ω(g(n))
Are there any functions such as f(n) and g(n) that both;
f(n) != O(g(n)) and
g(n) != O(f(n)).
Are there any functions that fulfills the requirements at the above?
f(n)=n and g(n)=n^(1 + sin(x)).
f(n) is not O(g(n)) and g(n) is not O(f(n)).
Refer http://c2.com/cgi/wiki?BigOh
Consider:
f(n) = 0 if n is odd, else n*n
g(n) = n
Then for odd values g(n) is more than a constant factor bigger than f(n) (and so g(n) is not O(f(n)), while for even values f(n) is more than a constant factor bigger than g(n) (and so f(n) is not O(g(n))).
Observe that f(n) does not have a limit at infinity as n approaches infinity, so in some sense this is a cheap example. But you could fix that by replacing 0, n, n*n with n, n*n, n*n*n.
I think if two non-negative functions have the property that f(n)/g(n) has a (perhaps infinite) limit as n approaches infinity, then it follows that one of them is big-O the other one. If the limit is 0 then f(n) is O(g(n)), if the limit is finite then each is big-O the other, and if the limit is infinite then g(n) is O(f(n)). But I'm too lazy to confirm by writing a proof.