Given a pseudocode like this:
int algorithm(int[] a, i, j)
if i > j
return 0
int n = j - i + 1
int d = n / 4
algorithm(a, i, i+3d-1)
algorithm(a, i+d, j)
return 1
The main cost is the cost given by the two recursive calls. The dimension of this two recursive calls should be in my opinion: 3n/4 for the first one (ignoring the 1) and n/4 for the second one. This means:
T(n) = T(3n/4) + T(n/4) + O(1)
Question 1: is this correct?
Question 2: is T(n) = 2T(3n/4) + O(1) correct?
In this code, the variables i and j are redundant because only their difference is used. So we can substitute j-i+1 with n everywhere. We can also ignore the argument a for clarity. The code becomes
int algorithm(n)
if 1>n
return 0
int d= n/4
algorithm(3d)
algorithm(n-1-d)
return 1
Next transformation: substitute d. We can also discard the returned value !
algorithm(n)
if 1>n
return
algorithm(3(n/4))
algorithm(n-1-n/4)
Related
static int isPascal(int n) {
int sum = 0;
int nthVal = 1;
while (sum < n) {
sum = sum + nthVal;
nthVal++;
}
return sum == n ? 1 : 0;
}
Here the function checks given number is pascal number or not. Pascal number is a number that is the sum of the integers from 1 to i for some i.
For example 6 is a Pascal number because 6 = 1 + 2 + 3
What will be the Time complexity of this function? Will it be O(logn) time? If so what will be base of log here?
If you consider calculating the square root a O(1) operation, you can do this check in O(1), with the help of the formula for the sum of the first i natural numbers
sum(i) = (i^2 + i)/2
Now in your case you don't know i but you know sum(i), because that's your n you want to check if it's a pascal number. So you have
n = (i^2 + i) /2
or
i^2 + i - 2n = 0
Solving this quadratic equation with the respecitive formula gives
i = -1/2 + sqrt(2*n + 1/4)
You can discard the second solution to this equation, because i must be > 0 to be a valid solution. If that resulting i is an integer, n is a Pascal number. Otherwise it isn't.
From that formula also follows, your iterative solution is in O(sqrt(n))
I need help with finding the complexity of a recursive algorith; I know that in order to solve this I have to find the linear recurrence, then apply the Master Theorem. As of my knowledge, finding the recurrence would be straightforward when only one parameter is considered;
In this case there are two parameters (i, j). Consider the function below called on (A,1,n):
integer stuff(integer [] A, integer i, integer j){
if i ≥ j then return i – j
integer h ← 0
for integer k ← 1 to floor((j – i + 1)/3) do {
h ← h + 1
}
return stuff(A, i , i + h) + stuff(A, j – h, j) – stuff(A, i + h + 1, j – h − 1)
}
Assuming various things, I guessed the relation to be:
T(1) = k
T(n) = T(n/3) + T(n/3) + T(n/3) + 1/3*n = 3*T(n/3) + 1/3*n
I assumed that because it looks that the function is called over 3 parts of 3, of which each is one third of n; being h = O(n/3)
First call: h+i-i = h ~ n/3
Second call: j-(j-h) = h ~ n/3
Third call: j-h-1-(i+h) = j-i-2h ~ n/3 (which I only assumed)
Even though I can try to guess the relation and make sense out of it, I don't know how to formally prove it.
If my guessing is correct, how do you get to that conclusion? If not, what am I missing?
Sorry for the long question, Thanks in advance
As you return inside the for, it means all the time the function will be finished just with a constant complexity! Because all the time goes to the for loop and it return the value of the function and everything is finished and the result is ready to be returned.
Also, the proof of the recurrent relationship comes from your analysis. If you use some counting principle in Combinatorics, the final result will be proved.
Moreover, if you correct the pseudocode and put the return at the end of the function, the complexity is T(n) = 3T(n/3) + \Theta(n) (as you analyzed). Now, from the master theorem, you can say that T(n) = n log(n)).
I am learning about time complexity and I trying to figure out a relationship. My lecture notes describe a recursive find functions as :
find(array A, item I)
if(arrayEmpty(A)) return BAD;
if(item == A[0]) return GOOD;
return find(allButFirst(A), I);
useful link : https://www.youtube.com/watch?v=_cG5KZSn1LE
My notes says that the starting relationship for time complexity is as a follow:
T(n) = 1 + T(n-1) // I understand this
T(1) = 1 // only one computation, understandable
then we unroll T(n)
T(N) = 1 + 1 + T(n-2) // every recursive step 1 comparison plus recursive call
T(N) = 1 + 1 + 1 + T(n-3)
T(N) = 1 + 1 + 1 + 1 + T(n-4)
...
T(N) = (n - 1) + T(n-(n-1)) // This point I am lost how they got this generalisation
If someone can explain how the above relation was generalised to T(N) = (n - 1) + T(n-(n-1)) and perhaps with a example would be better for clarity.
For example I want to try the above relationship with some values so let's say A {1, 2, 3} and I = 3
then here are the following computation
1 + T(n-1) // {2,3}
1 + 1 + T(n-2) // {3}
1 // {3} Found
So for above we had total 3 comparison and 2 recursive calls. So I would say the relationship is T(n) = n + t(n-(n-1)) = n + t(1) = n.
In each step of the unrolled pattern, if we are k steps into the recursion (for the first step, k=1, the last step, k=n), there are k 1's.
In the expansion you posted, the last line, T(N) = (n - 1) + T(n-(n-1)) is the second-to-last step, since T(n-(n-1)) expands to T(1), so for that line, k=n-1.
So there are n-1 1's in that line, hence the (n-1) term.
Likewise, the parameter passed to T at the kth step is n minus the current step, since it's the number of steps remaining. For the first row, that's n-k = n-1, hence T(n-1). For the second-to-last step, it's n-k = n-(n-1), hence T(n-(n-1)) = T(1).
I am given a formula f(n) where f(n) is defined, for all non-negative integers, as:
f(0) = 1
f(1) = 1
f(2) = 2
f(2n) = f(n) + f(n + 1) + n (for n > 1)
f(2n + 1) = f(n - 1) + f(n) + 1 (for n >= 1)
My goal is to find, for any given number s, the largest n where f(n) = s. If there is no such n return None. s can be up to 10^25.
I have a brute force solution using both recursion and dynamic programming, but neither is efficient enough. What concepts might help me find an efficient solution to this problem?
I want to add a little complexity analysis and estimate the size of f(n).
If you look at one recursive call of f(n), you notice, that the input n is basically divided by 2 before calling f(n) two times more, where always one call has an even and one has an odd input.
So the call tree is basically a binary tree where always the half of the nodes on a specific depth k provides a summand approx n/2k+1. The depth of the tree is log₂(n).
So the value of f(n) is in total about Θ(n/2 ⋅ log₂(n)).
Just to notice: This holds for even and odd inputs, but for even inputs the value is about an additional summand n/2 bigger. (I use Θ-notation to not have to think to much about some constants).
Now to the complexity:
Naive brute force
To calculate f(n) you have to call f(n) Θ(2log₂(n)) = Θ(n) times.
So if you want to calculate the values of f(n) until you reach s (or notice that there is no n with f(n)=s) you have to calculate f(n) s⋅log₂(s) times, which is in total Θ(s²⋅log(s)).
Dynamic programming
If you store every result of f(n), the time to calculate a f(n) reduces to Θ(1) (but it requires much more memory). So the total time complexity would reduce to Θ(s⋅log(s)).
Notice: Since we know f(n) ≤ f(n+2) for all n, you don't have to sort the values of f(n) and do a binary search.
Using binary search
Algorithm (input is s):
Set l = 1 and r = s
Set n = (l+r)/2 and round it to the next even number
calculate val = f(n).
if val == s then return n.
if val < s then set l = n
else set r = n.
goto 2
If you found a solution, fine. If not: try it again but round in step 2 to odd numbers. If this also does not return a solution, no solution exists at all.
This will take you Θ(log(s)) for the binary search and Θ(s) for the calculation of f(n) each time, so in total you get Θ(s⋅log(s)).
As you can see, this has the same complexity as the dynamic programming solution, but you don't have to save anything.
Notice: r = s does not hold for all s as an initial upper limit. However, if s is big enough, it holds. To be save, you can change the algorithm:
check first, if f(s) < s. If not, you can set l = s and r = 2s (or 2s+1 if it has to be odd).
Can you calculate the value of f(x) which x is from 0 to MAX_SIZE only once time?
what i mean is : calculate the value by DP.
f(0) = 1
f(1) = 1
f(2) = 2
f(3) = 3
f(4) = 7
f(5) = 4
... ...
f(MAX_SIZE) = ???
If the 1st step is illegal, exit. Otherwise, sort the value from small to big.
Such as 1,1,2,3,4,7,...
Now you can find whether exists n satisfied with f(n)=s in O(log(MAX_SIZE)) time.
Unfortunately, you don't mention how fast your algorithm should be. Perhaps you need to find some really clever rewrite of your formula to make it fast enough, in this case you might want to post this question on a mathematics forum.
The running time of your formula is O(n) for f(2n + 1) and O(n log n) for f(2n), according to the Master theorem, since:
T_even(n) = 2 * T(n / 2) + n / 2
T_odd(n) = 2 * T(n / 2) + 1
So the running time for the overall formula is O(n log n).
So if n is the answer to the problem, this algorithm would run in approx. O(n^2 log n), because you have to perform the formula roughly n times.
You can make this a little bit quicker by storing previous results, but of course, this is a tradeoff with memory.
Below is such a solution in Python.
D = {}
def f(n):
if n in D:
return D[n]
if n == 0 or n == 1:
return 1
if n == 2:
return 2
m = n // 2
if n % 2 == 0:
# f(2n) = f(n) + f(n + 1) + n (for n > 1)
y = f(m) + f(m + 1) + m
else:
# f(2n + 1) = f(n - 1) + f(n) + 1 (for n >= 1)
y = f(m - 1) + f(m) + 1
D[n] = y
return y
def find(s):
n = 0
y = 0
even_sol = None
while y < s:
y = f(n)
if y == s:
even_sol = n
break
n += 2
n = 1
y = 0
odd_sol = None
while y < s:
y = f(n)
if y == s:
odd_sol = n
break
n += 2
print(s,even_sol,odd_sol)
find(9992)
This recursive in every iteration for 2n and 2n+1 is increasing values, so if in any moment you will have value bigger, than s, then you can stop your algorithm.
To make effective algorithm you have to find or nice formula, that will calculate value, or make this in small loop, that will be much, much, much more effective, than your recursion. Your recursion is generally O(2^n), where loop is O(n).
This is how loop can be looking:
int[] values = new int[1000];
values[0] = 1;
values[1] = 1;
values[2] = 2;
for (int i = 3; i < values.length /2 - 1; i++) {
values[2 * i] = values[i] + values[i + 1] + i;
values[2 * i + 1] = values[i - 1] + values[i] + 1;
}
And inside this loop add condition of possible breaking it with success of failure.
I am somewhat confused with the running time analysis of a program here which has recursive calls which depend on a RNG. (Randomly Generated Number)
Let's begin with the pseudo-code, and then I will go into what I have thought about so far related to this one.
Func1(A, i, j)
/* A is an array of at least j integers */
1 if (i ≥ j) then return (0);
2 n ← j − i + 1 ; /* n = number of elements from i to j */
3 k ← Random(n);
4 s ← 0; //Takes time of Arbitrary C
5 for r ← i to j do
6 A[r] ← A[r] − A[i] − A[j]; //Arbitrary C
7 s ← s + A[r]; //Arbitrary C
8 end
9 s ← s + Func1(A, i, i+k-1); //Recursive Call 1
10 s ← s + Func1(A, i+k, j); //Recursive Call 2
11 return (s);
Okay, now let's get into the math I have tried so far. I'll try not to be too pedantic here as it is just a rough, estimated analysis of expected run time.
First, let's consider the worst case. Note that the K = Random(n) must be at least 1, and at most n. Therefore, the worst case is the K = 1 is picked. This causes the total running time to be equal to T(n) = cn + T(1) + T(n-1). Which means that overall it takes somewhere around cn^2 time total (you can use Wolfram to solve recurrence relations if you are stuck or rusty on recurrence relations, although this one is a fairly simple one).
Now, here is where I get somewhat confused. For the expected running time, we have to base our assumption off of the probability of the random number K. Therefore, we have to sum all the possible running times for different values of k, plus their individual probability. By lemma/hopefully intuitive logic: the probability of any one Randomly Generated k, with k between 1 to n, is equal 1/n.
Therefore, (in my opinion/analysis) the expected run time is:
ET(n) = cn + (1/n)*Summation(from k=1 to n-1) of (ET(k-1) + ET(n-k))
Let me explain a bit. The cn is simply for the loop which runs i to j. This is estimated by cn. The summation represents all of the possible values for k. The (1/n) multiplied by this summation is there because the probability of any one k is (1/n). The terms inside the summation represent the running times of the recursive calls of Func1. The first term on the left takes ET(k-1) because this recursive call is going to do a loop from i to k-1 (which is roughly ck), and then possibly call Func1 again. The second is a representation of the second recursive call, which would loop from i+k to j, which is also represented by n-k.
Upon expansion of the summation, we see that the overall function ET(n) is of the order n^2. However, as a test case, plugging in k=(n/2) gives a total running time for Func 1 of roughly nlog(n). This is why I am confused. How can this be, if the estimated running time is of the order n^2? Am I considering a "good" case by plugging in n/2 for k? Or am I thinking about k in the wrong sense in some way?
Expected time complexity is ET(n) = O(nlogn) . Following is math proof derived by myself please tell if any error :-
ET(n) = P(k=1)*(ET(1)+ET(n-1)) + P(k=2)*(ET(2)+ET(n-2)).......P(k=n-1)*(ET(n-1)+ET(1)) + c*n
As the RNG is uniformly random P(k=x) = 1/n for all x
hence ET(n) = 1/n*(ET(1)*2+ET(2)*2....ET(n-1)*2) + c*n
ET(n) = 2/n*sum(ET(i)) + c*n i in (1,n-1)
ET(n-1) = 2/(n-1)*sum(ET(i)) + c*(n-1) i in (1,n-2)
sum(ET(i)) i in (1,n-2) = (ET(n-1)-c*(n-1))*(n-1)/2
ET(n) = 2/n*(sum(ET(i)) in (1,n-2) + ET(n-1)) + c*n
ET(n) = 2/n*((ET(n-1)-c*(n-1))*(n-1)/2+ET(n-1)) + c*n
ET(n) = 2/n*((n+1)/2*ET(n-1) - c*(n-1)*(n-1)/2) + c*n
ET(n) = (n+1)/n*ET(n-1) + c*n - c*(n-1)*(n-1)/n
ET(n) = (n+1)/n*ET(n-1) + c
solving recurrence
ET(n) = (n+1)ET(1) + c + (n+1)/n*c + (n+1)/(n-1)*c + (n+1)/(n-2)*c.....
ET(n) = (n+1) + c + (n+1)*sum(1/i) i in (1,n)
sum(1/i) i in (1,n) = O(logn)
ET(n) = (n+1) + c + (n+1)*logn
ET(n) = O(nlogn)