Algorithms :XOR operation [duplicate] - algorithm

This question already has answers here:
Given an XOR and SUM of two numbers, how to find the number of pairs that satisfy them?
(2 answers)
Closed 5 years ago.
You are given a sum S and X , you need to find , if it there exist two numbers a and b such that a+b = S and a^b = X
I used a loop upto S/2 and check if it is possible or not
for(int i=0;i<=s/2;i++)
{
if(i^(s-i)==X)
return true;
}
complexity : O(n)
Need some better approach

Given that a+b = (a XOR b) + (a AND b)*2 (from here) we can calculate (a AND b):
If S < X => not possible, otherwise take S-X. If this is odd => not possible, otherwise (a AND b) = (S-X)/2.
now we can look at the bits of a and b individually. Checking all four combinations we see there is only one result that is impossible namely XOR and AND both 1.
So if (a XOR b) AND (a AND b) != 0 there is no solution. Otherwise one can find a and b that solve the equation.
if (S < X) return false;
Y = S - X;
if (Y is odd) return false;
if ((X & (Y/2)) != 0) return false;
return true;

Without previous knowledge of the equation, a+b = a^b + (a&b)*2, we can think of another solution. This solution is O(logK) where K is the maximum possible value of S and X. That is, if S and X are unsigned int then K is 2^32 - 1.
Start from the MSB of S and X. With the information that whether summing this bit must provide carry or not, we can check for this bit with condition that whether summing the bits to the right should provide carry or not.
Case1 ) summing must not provide carry
S X | need carry from right
------------------------------
0 0 | no (a = 0, b = 0)
0 1 | impossible
1 0 | yes (a = 0, b = 0)
1 1 | no (a = 1, b = 0 or 0,1)
Case2 ) summing must provide carry
S X | need carry from right
------------------------------
0 0 | no (a = 1, b = 1)
0 1 | yes (a = 1, b = 0 or 0,1)
1 0 | yes (a = 1, b = 1)
1 1 | impossible
There is a special case for the MSB where the carry doesn't matter.
Case3 ) don't care
S X | need carry from right
------------------------------
0 0 | no (a = 0, b = 0 or 1,1)
0 1 | yes (a = 1, b = 0 or 0,1)
1 0 | yes (a = 0, b = 0 or 1,1)
1 1 | no (a = 1, b = 0 or 0,1)
Lastly, the LSB must end with 'no need carry from right'.
The implementation and test code is here. It compares the output of the accepted solution and this solution.

I assume that you are assuming that your input sequence is sorted.
If it is, then this problem is as good as finding pair for a given sum and while checking sum check for their (a^b == SUM) this should be enough and this problem can be solved in O(n). URL
And you can't do better than that. In worst case you have to visit each element atleast once.

Related

find a Xor b equals to n.can you suggest me the efficient way of finding xor value

let there is two number a and b such that a^b is n.I have given n value I need to find a and b.
EXAMPLE
n=5
a=1 b=4 and a^b=5
similarly a=9 b=12
please help to find this.'
If there is no constrain on value of a and b then a can be 0 and b is equal to n (because xor of any number with zero is equal to same number).
If 0<a,b<n then you can iterate from 1 to n , and assign that value to a ,then value of b will be n^a check if b is less than n or not , if it is then print a and b
int a,b,n;
a=-1;
b=-1;
cin>>n;
for(int i=1;i<=n;i++)
{
b=i^n;
if((b>0) && (b<=n))
{
a=i;
break;
}
}
cout<<a<<" "<<b;
Write both numbers in binary form and compute the XOR bit-by-bit:
a = 9 1 0 0 1
b = 12 1 1 0 0
a ^ b = n 0 1 0 1 decimal 5
a ^ n = b 1 1 0 0
b ^ n = a 1 0 0 1
In general terms:
a w x y z
b w !x y !z
n 0 1 0 1
You can take an arbitrary a, invert two bits as shown to get b such that a ^ b = 5.

Finding the largest power of a number that divides a factorial in haskell

So I am writing a haskell program to calculate the largest power of a number that divides a factorial.
largestPower :: Int -> Int -> Int
Here largestPower a b has find largest power of b that divides a!.
Now I understand the math behind it, the way to find the answer is to repeatedly divide a (just a) by b, ignore the remainder and finally add all the quotients. So if we have something like
largestPower 10 2
we should get 8 because 10/2=5/2=2/2=1 and we add 5+2+1=8
However, I am unable to figure out how to implement this as a function, do I use arrays or just a simple recursive function.
I am gravitating towards it being just a normal function, though I guess it can be done by storing quotients in an array and adding them.
Recursion without an accumulator
You can simply write a recursive algorithm and sum up the result of each call. Here we have two cases:
a is less than b, in which case the largest power is 0. So:
largestPower a b | a < b = 0
a is greater than or equal to b, in that case we divide a by b, calculate largestPower for that division, and add the division to the result. Like:
| otherwise = d + largestPower d b
where d = (div a b)
Or putting it together:
largestPower a b | a < b = 1
| otherwise = d + largestPower d b
where d = (div a b)
Recursion with an accumuator
You can also use recursion with an accumulator: a variable you pass through the recursion, and update accordingly. At the end, you return that accumulator (or a function called on that accumulator).
Here the accumulator would of course be the running product of divisions, so:
largestPower = largestPower' 0
So we will define a function largestPower' (mind the accent) with an accumulator as first argument that is initialized as 1.
Now in the recursion, there are two cases:
a is less than b, we simply return the accumulator:
largestPower' r a b | a < b = r
otherwise we multiply our accumulator with b, and pass the division to the largestPower' with a recursive call:
| otherwise = largestPower' (d+r) d b
where d = (div a b)
Or the full version:
largestPower = largestPower' 1
largestPower' r a b | a < b = r
| otherwise = largestPower' (d+r) d b
where d = (div a b)
Naive correct algorithm
The algorithm is not correct. A "naive" algorithm would be to simply divide every item and keep decrementing until you reach 1, like:
largestPower 1 _ = 0
largestPower a b = sumPower a + largestPower (a-1) b
where sumPower n | n `mod` b == 0 = 1 + sumPower (div n b)
| otherwise = 0
So this means that for the largestPower 4 2, this can be written as:
largestPower 4 2 = sumPower 4 + sumPower 3 + sumPower 2
and:
sumPower 4 = 1 + sumPower 2
= 1 + 1 + sumPower 1
= 1 + 1 + 0
= 2
sumPower 3 = 0
sumPower 2 = 1 + sumPower 1
= 1 + 0
= 1
So 3.
The algorithm as stated can be implemented quite simply:
largestPower :: Int -> Int -> Int
largestPower 0 b = 0
largestPower a b = d + largestPower d b where d = a `div` b
However, the algorithm is not correct for composite b. For example, largestPower 10 6 with this algorithm yields 1, but in fact the correct answer is 4. The problem is that this algorithm ignores multiples of 2 and 3 that are not multiples of 6. How you fix the algorithm is a completely separate question, though.

How does less than and greater than work on a logical level in binary?

I see how equality could work by just comparing bit patterns, but how would one write one's own less than and greater than operators? What is the actual process to mechanically compare values to each other?
I assume you want to just know how logic does it? And mimic that? Well here goes:
First off you have to talk about unsigned greater than or less than vs signed greater than vs less than because they matter. Just do three bit numbers to make life easier, it all scales up to N bits wide.
As processor documentation usually states a compare instruction will do a subtraction of the two operands in order to generate flags, so it is a subtract that doesnt save the answer just modifies the flags. That then later jump or branch on some flag pattern can be used. Some processors dont use flags but have a similar solution, they still use the equivalent of flags but just dont save them anywhere. compare and branch if greater than kind of an instruction instead of separate compare and branch if greater than instructions.
What is a subtract in logic? Well in grade school we learned that
a - b = a + (-b).
We also know from intro programming classes that a negative in twos complement means invert and add one so
a - b = a + (~b) + 1.
Note ones complement means to invert ~b in C means invert all the bits it is also known as taking the ones complement.
so 7 - 5 is
1
111
+010
=====
Pretty cool we can use the carry in as the "plus one".
1111
111
+010
=====
010
So the answer is 2 with the carry out set. Carry out set is a good thing means we didnt borrow. Not all processors do the same but so far with a subtract we use an adder but invert the second operand and invert the carry in. If we invert the carry out we can call it a borrow, 7 - 5 doesnt borrow. Again some processor architectures dont invert and call it a borrow they just call it carry out. It still goes into the carry flag if they have flags.
Why does any of this matter? Just hang on.
Lets look at 6-5 5-5 and 4-5 and see what the flags tell us.
1101
110
+010
=====
001
1111
101
+010
=====
000
0001
100
+010
=====
111
So what this means is the carry out tells us (unsigned) less than if 0. If if 1 then greater than or equal. That was using a - b where b is what we were comparing against. So if we then do b - a that would imply that we can get (unsigned) greater than with the carry bit. 5 - 4, 5 - 5, 5 - 6. We already know what 5 - 5 looks like zero with the carry out set
1111
101
011
====
001
0011
101
001
====
111
Yep we can determine (unsigned) greater than or equal or (unsigned) less than or equal using the carry flag. Not less than is the same as greater than or equal and vice versa. You just need to get the operand you want to compare against in the right place. I probably did this backward from every compare instruction as I think they subtract a - b where a is what is being compared against.
Now that you have seen this you can easily scratch out the above math on a piece of paper in a few seconds to know which order you need things and what flags to use. Or do a simple experiment like this with a processor you are using and look at the flags to figure out which is subtracted from which and/or whether or not they invert the carry out and call it a borrow.
We can see from doing it with paper and pencil as in grade school that addition boils down to one column at a time, you have a carry in plus two operands, and you get a result and a carry out. You can cascade the carry out to the carry in of the next column and repeat this for any number of bits you can afford to store.
Some/many instruction sets use flags (carry, zero, signed overflow, and negative are the set you need to do most comparisons). You can see I hope that you dont need assembly language nor an instruction set with flags, you can do this yourself with a programming language that has the basic boolean and math operations.
Untested code I just banged out in this edit window.
unsigned char a;
unsigned char b;
unsigned char aa;
unsigned char bb;
unsigned char c;
aa = a&0xF;
bb = (~b)&0xF;
c = aa + bb + 1;
c >>=4;
a >>=4;
b >>=4;
aa = a&0xF;
bb = (~b)&0xF;
c = c + aa + bb;
c >>=4;
And there you go, using an equals comparison, compare c with zero. And depending on your operand order it tells you (unsigned) less than or greater than. If you want to do more than 8 bits then keep on cascading that add with carry indefinitely.
Signed numbers...
ADDITION (and subtraction) LOGIC DOES NOT KNOW THE DIFFERENCE BETWEEN SIGNED AND UNSIGNED OPERANDS. Important to know. This is the beauty of twos complement. Try it yourself write a program that adds bit patterns together and prints out the bit patterns. Interpret those bit patterns input and output as all unsigned or all signed and you see this works (note some combinations overflow and the result is clipped).
Now saying that the flags do vary for subtraction comparisons. We know from grade school math where we carry the one or whatever over as we see in binary as well. The carry out is an unsigned overflow for unsigned. If set you overflowed we have a 1 we cant fit into our register so the result is too big we fail. Signed overflow though is the V bit which tells us if the carry IN and carry OUT of the msbit were the same or not.
Now lets use four bits, cause I want to. We can do the 5 - 4, 5 - 5, and 5 - 6. These are positive numbers so we have seen this but we didnt look at the V flag nor N flag (nor z flag). The N flag is the msbit of the result which indicates negative using twos complement notation, not to be confused with a sign bit although it is as a side effect, it is just not a separate sign bit that you remove from the number.
11111
0101
1011
=====
0001
c = 1, v = 0, n = 0
11111
0101
1010
=====
0000
c = 1, v = 0, n = 0
00011
0101
1001
=====
1111
c = 0, v = 0, n = 1
Now negative numbers -5 - -6, -5 - -5, -5 - -4
10111
1011
1010
=====
0110
c = 0, v = 1, n = 1
You know what, there is an easier way.
#include <stdio.h>
int main ( void )
{
unsigned int ra;
unsigned int rb;
unsigned int rc;
unsigned int rd;
unsigned int re;
int a;
int b;
for(a=-5;a<=5;a++)
{
for(b=-5;b<=5;b++)
{
ra = a&0xF;
rb = (-b)&0xF;
rc = ra+rb;
re = rc&8;
re >>=3;
rc >>=4;
ra = a&0x7;
rb = (-b)&0x7;
rd = ra+rb;
rd >>=3;
rd += rc;
rd &=1;
printf("%+d vs %+d: c = %u, n = %u, v = %u\n",a,b,rc,re,rd);
}
}
return(0);
}
and a subset of the results
-5 vs -5: c = 1, n = 0, v = 0
-5 vs -4: c = 0, n = 1, v = 0
-4 vs -5: c = 1, n = 0, v = 0
-4 vs -4: c = 1, n = 0, v = 0
-4 vs -3: c = 0, n = 1, v = 0
-3 vs -4: c = 1, n = 0, v = 0
-3 vs -3: c = 1, n = 0, v = 0
-3 vs -2: c = 0, n = 1, v = 0
-2 vs -3: c = 1, n = 0, v = 0
-2 vs -2: c = 1, n = 0, v = 0
-2 vs -1: c = 0, n = 1, v = 0
-1 vs -2: c = 1, n = 0, v = 0
-1 vs -1: c = 1, n = 0, v = 0
-1 vs +0: c = 0, n = 1, v = 0
+0 vs -1: c = 0, n = 0, v = 0
+0 vs +0: c = 0, n = 0, v = 0
+0 vs +1: c = 0, n = 1, v = 0
+1 vs +0: c = 0, n = 0, v = 0
+1 vs +1: c = 1, n = 0, v = 0
+1 vs +2: c = 0, n = 1, v = 0
+3 vs +2: c = 1, n = 0, v = 0
+3 vs +3: c = 1, n = 0, v = 0
+3 vs +4: c = 0, n = 1, v = 0
And I will just tell you the answer...You are looking for n == v or not n == v. So if you compute n and v, then x = (n+v)&1. Then if that is a zero they were equal, if that is a 1 they were not. You can use an equals comparison. When they were not equal b was greater than a. Reverse your operands and you can check b less than a.
You can change the code above to only print out if n and v are equal. So if you are using a processor with only an equals comparison, you can still survive with real programming languages and comparisons.
Some processor manuals may chart this out for you. They may say n==v for one thing or not z and n!=v for the other (LT vs GT). But it could be simplified, from grade school the alligator eats the bigger one a > b when you flip it b < a. So feed the operators in one way and you get a > b, feed them the other way you get b < a.
Equals is just a straight bit comparison that goes through a separate logic path, not something that falls out of the addition. N is grabbed from the msbit of the result. C and V fall out of the addition.
One method that is O(1) over all numbers in the given domain (keeping the number of bits constant) would be to subtract the numbers and check the sign bit. I.e., suppose you have A and B, then
C = A - B
If C's sign bit is 1, then B > A. Otherwise, A >= B.

Algorithm for checking if number is factorable into a set of prime numbers

I was wondering if there is an algorithm that checks wether a given number is factorable into a set of prime numbers and if no give out the nearest number.
The problem comes always up when I use the FFT.
Thanks a lot for your help guys.
In general this looks like a hard problem, particularly finding the next largest integer that factors into your set of primes. However, if your set of primes isn't too big, one approach would be to turn this into an integer optimization problem by taking the logs. Here is how to find the smallest number > n that factors into a set of primes p_1...p_k
choose integers x_1,...,x_k to minimize (x_1 log p_1 + ... + x_k log p_k - log n)
Subject to:
x_1 log p_1 + ... + x_k log p_k >= log n
x_i >= 0 for all i
The x_i will give you the exponents for the primes. Here is an implementation in R using lpSolve:
minfact<-function(x,p){
sol<-lp("min",log(p),t(log(p)),">=",log(x),all.int=T)
prod(p^sol$solution)
}
> p<-c(2,3,13,31)
> x<-124363183
> y<-minfact(x,p)
> y
[1] 124730112
> factorize(y)
Big Integer ('bigz') object of length 13:
[1] 2 2 2 2 2 2 2 2 3 13 13 31 31
> y-x
[1] 366929
>
Using big integers, this works pretty well even for large numbers:
> p<-c(2,3,13,31,53,79)
> x<-as.bigz("1243631831278461278641361")
> y<-minfact(x,p)
y
>
Big Integer ('bigz') :
[1] 1243634072805560436129792
> factorize(y)
Big Integer ('bigz') object of length 45:
[1] 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
[26] 2 2 2 2 2 2 2 2 3 3 3 3 13 31 31 31 31 53 53 53
>
Your question is about well-known factorization problem - which could not be resolved with 'fast' (polynomial) time. Lenstra's elliptic algorithm is the most efficient (known) way in common case, but it requires strong knowledge of numbers theory - and it's also sub-exponential (but not polynomial).
Other algorithms are listed in the page by first link in my post, but such things as direct try (brute force) are much more slower, of cause.
Please, note, that under "could not be resolved with polynomial time" - I mean that there's no way to do this now - but not that such way does not exist (at least now, number theory can not provide such solution for this problem)
Here is a brute force method in C++. It returns the factorization of the nearest factorable number. If N has two equidistant factorable neighbours, it returns the smallest one.
GCC 4.7.3: g++ -Wall -Wextra -std=c++0x factorable-neighbour.cpp
#include <iostream>
#include <vector>
using ints = std::vector<int>;
ints factor(int n, const ints& primes) {
ints f(primes.size(), 0);
for (int i = 0; i < primes.size(); ++i) {
while (0< n && !(n % primes[i])) {
n /= primes[i];
++f[i]; } }
// append the "remainder"
f.push_back(n);
return f;
}
ints closest_factorable(int n, const ints& primes) {
int d = 0;
ints r;
while (true) {
r = factor(n + d, primes);
if (r[r.size() - 1] == 1) { break; }
++d;
r = factor(n - d, primes);
if (r[r.size() - 1] == 1) { break; }
}
r.pop_back();
return r; }
int main() {
for (int i = 0; i < 30; ++i) {
for (const auto& f : closest_factorable(i, {2, 3, 5, 7, 11})) {
std::cout << f << " "; }
std::cout << "\n"; }
}
I suppose that you have a (small) set of prime numbers S and an integer n and you want to know is n factors only using number in S. The easiest way seems to be the following:
P <- product of s in S
while P != 1 do
P <- GCD(P, n)
n <- n/P
return n == 1
You compute the GCD using Euclid's algorithm.
The idea is the following: Suppose that S = {p1, p2, ... ,pk}. You can write n uniquely as
n = p1^n1 p2^n2 ... pk^nk * R
where R is coprime wrt the pi. You want to know whether R=1.
Then
GCD(n, P) = prod ( pi such that ni <> 0 ).
Therefore n/p decrease all those non zeros ni by 1 so that they eventually become 0. At the end only R remains.
For example: S = {2,3,5}, n = 5600 = 2^5*5^2*7. Then P = 2*3*5 = 30. One gets GCD(n, p)=10=2*5. And therefore n/GCD(n,p) = 560 = 2^4*5*7.
You are now back to the same problem: You want to know if 560 can be factored using S = {2,5} hence the loop. So the next steps are
GCD(560, 10) = 10. 560/GCD = 56 = 2^3 * 7.
GCD(56, 10) = 2. 56/2 = 28 = 2^2 * 7
GCD(28, 2) = 2. 28/2 = 14 = 2 * 7
GCD(14, 2) = 2. 14/2 = 7
GCD(7, 2) = 1 so that R = 7 ! Your answer if FALSE.
kissfft has a function
int kiss_fft_next_fast_size(int n)
that returns the next largest N that is an aggregate of 2,3,5.
Also related is a kf_factor function that factorizes a number n, pulling out the "nice" FFT primes first (e.g. 4's are pulled out before 2's)

nᵗʰ ugly number

Numbers whose only prime factors are 2, 3, or 5 are called ugly numbers.
Example:
1, 2, 3, 4, 5, 6, 8, 9, 10, 12, 15, ...
1 can be considered as 2^0.
I am working on finding nth ugly number. Note that these numbers are extremely sparsely distributed as n gets large.
I wrote a trivial program that computes if a given number is ugly or not. For n > 500 - it became super slow. I tried using memoization - observation: ugly_number * 2, ugly_number * 3, ugly_number * 5 are all ugly. Even with that it is slow. I tried using some properties of log - since that will reduce this problem from multiplication to addition - but, not much luck yet. Thought of sharing this with you all. Any interesting ideas?
Using a concept similar to Sieve of Eratosthenes (thanks Anon)
for (int i(2), uglyCount(0); ; i++) {
if (i % 2 == 0)
continue;
if (i % 3 == 0)
continue;
if (i % 5 == 0)
continue;
uglyCount++;
if (uglyCount == n - 1)
break;
}
i is the nth ugly number.
Even this is pretty slow. I am trying to find the 1500th ugly number.
A simple fast solution in Java. Uses approach described by Anon..
Here TreeSet is just a container capable of returning smallest element in it. (No duplicates stored.)
int n = 20;
SortedSet<Long> next = new TreeSet<Long>();
next.add((long) 1);
long cur = 0;
for (int i = 0; i < n; ++i) {
cur = next.first();
System.out.println("number " + (i + 1) + ": " + cur);
next.add(cur * 2);
next.add(cur * 3);
next.add(cur * 5);
next.remove(cur);
}
Since 1000th ugly number is 51200000, storing them in bool[] isn't really an option.
edit
As a recreation from work (debugging stupid Hibernate), here's completely linear solution. Thanks to marcog for idea!
int n = 1000;
int last2 = 0;
int last3 = 0;
int last5 = 0;
long[] result = new long[n];
result[0] = 1;
for (int i = 1; i < n; ++i) {
long prev = result[i - 1];
while (result[last2] * 2 <= prev) {
++last2;
}
while (result[last3] * 3 <= prev) {
++last3;
}
while (result[last5] * 5 <= prev) {
++last5;
}
long candidate1 = result[last2] * 2;
long candidate2 = result[last3] * 3;
long candidate3 = result[last5] * 5;
result[i] = Math.min(candidate1, Math.min(candidate2, candidate3));
}
System.out.println(result[n - 1]);
The idea is that to calculate a[i], we can use a[j]*2 for some j < i. But we also need to make sure that 1) a[j]*2 > a[i - 1] and 2) j is smallest possible.
Then, a[i] = min(a[j]*2, a[k]*3, a[t]*5).
I am working on finding nth ugly number. Note that these numbers are extremely sparsely distributed as n gets large.
I wrote a trivial program that computes if a given number is ugly or not.
This looks like the wrong approach for the problem you're trying to solve - it's a bit of a shlemiel algorithm.
Are you familiar with the Sieve of Eratosthenes algorithm for finding primes? Something similar (exploiting the knowledge that every ugly number is 2, 3 or 5 times another ugly number) would probably work better for solving this.
With the comparison to the Sieve I don't mean "keep an array of bools and eliminate possibilities as you go up". I am more referring to the general method of generating solutions based on previous results. Where the Sieve gets a number and then removes all multiples of it from the candidate set, a good algorithm for this problem would start with an empty set and then add the correct multiples of each ugly number to that.
My answer refers to the correct answer given by Nikita Rybak.
So that one could see a transition from the idea of the first approach to that of the second.
from collections import deque
def hamming():
h=1;next2,next3,next5=deque([]),deque([]),deque([])
while True:
yield h
next2.append(2*h)
next3.append(3*h)
next5.append(5*h)
h=min(next2[0],next3[0],next5[0])
if h == next2[0]: next2.popleft()
if h == next3[0]: next3.popleft()
if h == next5[0]: next5.popleft()
What's changed from Nikita Rybak's 1st approach is that, instead of adding next candidates into single data structure, i.e. Tree set, one can add each of them separately into 3 FIFO lists. This way, each list will be kept sorted all the time, and the next least candidate must always be at the head of one ore more of these lists.
If we eliminate the use of the three lists above, we arrive at the second implementation in Nikita Rybak' answer. This is done by evaluating those candidates (to be contained in three lists) only when needed, so that there is no need to store them.
Simply put:
In the first approach, we put every new candidate into single data structure, and that's bad because too many things get mixed up unwisely. This poor strategy inevitably entails O(log(tree size)) time complexity every time we make a query to the structure. By putting them into separate queues, however, you will see that each query takes only O(1) and that's why the overall performance reduces to O(n)!!! This is because each of the three lists is already sorted, by itself.
I believe you can solve this problem in sub-linear time, probably O(n^{2/3}).
To give you the idea, if you simplify the problem to allow factors of just 2 and 3, you can achieve O(n^{1/2}) time starting by searching for the smallest power of two that is at least as large as the nth ugly number, and then generating a list of O(n^{1/2}) candidates. This code should give you an idea how to do it. It relies on the fact that the nth number containing only powers of 2 and 3 has a prime factorization whose sum of exponents is O(n^{1/2}).
def foo(n):
p2 = 1 # current power of 2
p3 = 1 # current power of 3
e3 = 0 # exponent of current power of 3
t = 1 # number less than or equal to the current power of 2
while t < n:
p2 *= 2
if p3 * 3 < p2:
p3 *= 3
e3 += 1
t += 1 + e3
candidates = [p2]
c = p2
for i in range(e3):
c /= 2
c *= 3
if c > p2: c /= 2
candidates.append(c)
return sorted(candidates)[n - (t - len(candidates))]
The same idea should work for three allowed factors, but the code gets more complex. The sum of the powers of the factorization drops to O(n^{1/3}), but you need to consider more candidates, O(n^{2/3}) to be more precise.
A lot of good answers here, but I was having trouble understanding those, specifically how any of these answers, including the accepted one, maintained the axiom 2 in Dijkstra's original paper:
Axiom 2. If x is in the sequence, so is 2 * x, 3 * x, and 5 * x.
After some whiteboarding, it became clear that the axiom 2 is not an invariant at each iteration of the algorithm, but actually the goal of the algorithm itself. At each iteration, we try to restore the condition in axiom 2. If last is the last value in the result sequence S, axiom 2 can simply be rephrased as:
For some x in S, the next value in S is the minimum of 2x,
3x, and 5x, that is greater than last. Let's call this axiom 2'.
Thus, if we can find x, we can compute the minimum of 2x, 3x, and 5x in constant time, and add it to S.
But how do we find x? One approach is, we don't; instead, whenever we add a new element e to S, we compute 2e, 3e, and 5e, and add them to a minimum priority queue. Since this operations guarantees e is in S, simply extracting the top element of the PQ satisfies axiom 2'.
This approach works, but the problem is that we generate a bunch of numbers we may not end up using. See this answer for an example; if the user wants the 5th element in S (5), the PQ at that moment holds 6 6 8 9 10 10 12 15 15 20 25. Can we not waste this space?
Turns out, we can do better. Instead of storing all these numbers, we simply maintain three counters for each of the multiples, namely, 2i, 3j, and 5k. These are candidates for the next number in S. When we pick one of them, we increment only the corresponding counter, and not the other two. By doing so, we are not eagerly generating all the multiples, thus solving the space problem with the first approach.
Let's see a dry run for n = 8, i.e. the number 9. We start with 1, as stated by axiom 1 in Dijkstra's paper.
+---------+---+---+---+----+----+----+-------------------+
| # | i | j | k | 2i | 3j | 5k | S |
+---------+---+---+---+----+----+----+-------------------+
| initial | 1 | 1 | 1 | 2 | 3 | 5 | {1} |
+---------+---+---+---+----+----+----+-------------------+
| 1 | 1 | 1 | 1 | 2 | 3 | 5 | {1,2} |
+---------+---+---+---+----+----+----+-------------------+
| 2 | 2 | 1 | 1 | 4 | 3 | 5 | {1,2,3} |
+---------+---+---+---+----+----+----+-------------------+
| 3 | 2 | 2 | 1 | 4 | 6 | 5 | {1,2,3,4} |
+---------+---+---+---+----+----+----+-------------------+
| 4 | 3 | 2 | 1 | 6 | 6 | 5 | {1,2,3,4,5} |
+---------+---+---+---+----+----+----+-------------------+
| 5 | 3 | 2 | 2 | 6 | 6 | 10 | {1,2,3,4,5,6} |
+---------+---+---+---+----+----+----+-------------------+
| 6 | 4 | 2 | 2 | 8 | 6 | 10 | {1,2,3,4,5,6} |
+---------+---+---+---+----+----+----+-------------------+
| 7 | 4 | 3 | 2 | 8 | 9 | 10 | {1,2,3,4,5,6,8} |
+---------+---+---+---+----+----+----+-------------------+
| 8 | 5 | 3 | 2 | 10 | 9 | 10 | {1,2,3,4,5,6,8,9} |
+---------+---+---+---+----+----+----+-------------------+
Notice that S didn't grow at iteration 6, because the minimum candidate 6 had already been added previously. To avoid this problem of having to remember all of the previous elements, we amend our algorithm to increment all the counters whenever the corresponding multiples are equal to the minimum candidate. That brings us to the following Scala implementation.
def hamming(n: Int): Seq[BigInt] = {
#tailrec
def next(x: Int, factor: Int, xs: IndexedSeq[BigInt]): Int = {
val leq = factor * xs(x) <= xs.last
if (leq) next(x + 1, factor, xs)
else x
}
#tailrec
def loop(i: Int, j: Int, k: Int, xs: IndexedSeq[BigInt]): IndexedSeq[BigInt] = {
if (xs.size < n) {
val a = next(i, 2, xs)
val b = next(j, 3, xs)
val c = next(k, 5, xs)
val m = Seq(2 * xs(a), 3 * xs(b), 5 * xs(c)).min
val x = a + (if (2 * xs(a) == m) 1 else 0)
val y = b + (if (3 * xs(b) == m) 1 else 0)
val z = c + (if (5 * xs(c) == m) 1 else 0)
loop(x, y, z, xs :+ m)
} else xs
}
loop(0, 0, 0, IndexedSeq(BigInt(1)))
}
Basicly the search could be made O(n):
Consider that you keep a partial history of ugly numbers. Now, at each step you have to find the next one. It should be equal to a number from the history multiplied by 2, 3 or 5. Chose the smallest of them, add it to history, and drop some numbers from it so that the smallest from the list multiplied by 5 would be larger than the largest.
It will be fast, because the search of the next number will be simple:
min(largest * 2, smallest * 5, one from the middle * 3),
that is larger than the largest number in the list. If they are scarse, the list will always contain few numbers, so the search of the number that have to be multiplied by 3 will be fast.
Here is a correct solution in ML. The function ugly() will return a stream (lazy list) of hamming numbers. The function nth can be used on this stream.
This uses the Sieve method, the next elements are only calculated when needed.
datatype stream = Item of int * (unit->stream);
fun cons (x,xs) = Item(x, xs);
fun head (Item(i,xf)) = i;
fun tail (Item(i,xf)) = xf();
fun maps f xs = cons(f (head xs), fn()=> maps f (tail xs));
fun nth(s,1)=head(s)
| nth(s,n)=nth(tail(s),n-1);
fun merge(xs,ys)=if (head xs=head ys) then
cons(head xs,fn()=>merge(tail xs,tail ys))
else if (head xs<head ys) then
cons(head xs,fn()=>merge(tail xs,ys))
else
cons(head ys,fn()=>merge(xs,tail ys));
fun double n=n*2;
fun triple n=n*3;
fun ij()=
cons(1,fn()=>
merge(maps double (ij()),maps triple (ij())));
fun quint n=n*5;
fun ugly()=
cons(1,fn()=>
merge((tail (ij())),maps quint (ugly())));
This was first year CS work :-)
To find the n-th ugly number in O (n^(2/3)), jonderry's algorithm will work just fine. Note that the numbers involved are huge so any algorithm trying to check whether a number is ugly or not has no chance.
Finding all of the n smallest ugly numbers in ascending order is done easily by using a priority queue in O (n log n) time and O (n) space: Create a priority queue of numbers with the smallest numbers first, initially including just the number 1. Then repeat n times: Remove the smallest number x from the priority queue. If x hasn't been removed before, then x is the next larger ugly number, and we add 2x, 3x and 5x to the priority queue. (If anyone doesn't know the term priority queue, it's like the heap in the heapsort algorithm). Here's the start of the algorithm:
1 -> 2 3 5
1 2 -> 3 4 5 6 10
1 2 3 -> 4 5 6 6 9 10 15
1 2 3 4 -> 5 6 6 8 9 10 12 15 20
1 2 3 4 5 -> 6 6 8 9 10 10 12 15 15 20 25
1 2 3 4 5 6 -> 6 8 9 10 10 12 12 15 15 18 20 25 30
1 2 3 4 5 6 -> 8 9 10 10 12 12 15 15 18 20 25 30
1 2 3 4 5 6 8 -> 9 10 10 12 12 15 15 16 18 20 24 25 30 40
Proof of execution time: We extract an ugly number from the queue n times. We initially have one element in the queue, and after extracting an ugly number we add three elements, increasing the number by 2. So after n ugly numbers are found we have at most 2n + 1 elements in the queue. Extracting an element can be done in logarithmic time. We extract more numbers than just the ugly numbers but at most n ugly numbers plus 2n - 1 other numbers (those that could have been in the sieve after n-1 steps). So the total time is less than 3n item removals in logarithmic time = O (n log n), and the total space is at most 2n + 1 elements = O (n).
I guess we can use Dynamic Programming (DP) and compute nth Ugly Number. Complete explanation can be found at http://www.geeksforgeeks.org/ugly-numbers/
#include <iostream>
#define MAX 1000
using namespace std;
// Find Minimum among three numbers
long int min(long int x, long int y, long int z) {
if(x<=y) {
if(x<=z) {
return x;
} else {
return z;
}
} else {
if(y<=z) {
return y;
} else {
return z;
}
}
}
// Actual Method that computes all Ugly Numbers till the required range
long int uglyNumber(int count) {
long int arr[MAX], val;
// index of last multiple of 2 --> i2
// index of last multiple of 3 --> i3
// index of last multiple of 5 --> i5
int i2, i3, i5, lastIndex;
arr[0] = 1;
i2 = i3 = i5 = 0;
lastIndex = 1;
while(lastIndex<=count-1) {
val = min(2*arr[i2], 3*arr[i3], 5*arr[i5]);
arr[lastIndex] = val;
lastIndex++;
if(val == 2*arr[i2]) {
i2++;
}
if(val == 3*arr[i3]) {
i3++;
}
if(val == 5*arr[i5]) {
i5++;
}
}
return arr[lastIndex-1];
}
// Starting point of program
int main() {
long int num;
int count;
cout<<"Which Ugly Number : ";
cin>>count;
num = uglyNumber(count);
cout<<endl<<num;
return 0;
}
We can see that its quite fast, just change the value of MAX to compute higher Ugly Number
Using 3 generators in parallel and selecting the smallest at each iteration, here is a C program to compute all ugly numbers below 2128 in less than 1 second:
#include <limits.h>
#include <stdio.h>
#if 0
typedef unsigned long long ugly_t;
#define UGLY_MAX (~(ugly_t)0)
#else
typedef __uint128_t ugly_t;
#define UGLY_MAX (~(ugly_t)0)
#endif
int print_ugly(int i, ugly_t u) {
char buf[64], *p = buf + sizeof(buf);
*--p = '\0';
do { *--p = '0' + u % 10; } while ((u /= 10) != 0);
return printf("%d: %s\n", i, p);
}
int main() {
int i = 0, n2 = 0, n3 = 0, n5 = 0;
ugly_t u, ug2 = 1, ug3 = 1, ug5 = 1;
#define UGLY_COUNT 110000
ugly_t ugly[UGLY_COUNT];
while (i < UGLY_COUNT) {
u = ug2;
if (u > ug3) u = ug3;
if (u > ug5) u = ug5;
if (u == UGLY_MAX)
break;
ugly[i++] = u;
print_ugly(i, u);
if (u == ug2) {
if (ugly[n2] <= UGLY_MAX / 2)
ug2 = 2 * ugly[n2++];
else
ug2 = UGLY_MAX;
}
if (u == ug3) {
if (ugly[n3] <= UGLY_MAX / 3)
ug3 = 3 * ugly[n3++];
else
ug3 = UGLY_MAX;
}
if (u == ug5) {
if (ugly[n5] <= UGLY_MAX / 5)
ug5 = 5 * ugly[n5++];
else
ug5 = UGLY_MAX;
}
}
return 0;
}
Here are the last 10 lines of output:
100517: 338915443777200000000000000000000000000
100518: 339129266201729628114355465608000000000
100519: 339186548067800934969350553600000000000
100520: 339298130282929870605468750000000000000
100521: 339467078447341918945312500000000000000
100522: 339569540691046437734055936000000000000
100523: 339738624000000000000000000000000000000
100524: 339952965770562084651663360000000000000
100525: 340010386766614455386112000000000000000
100526: 340122240000000000000000000000000000000
Here is a version in Javascript usable with QuickJS:
import * as std from "std";
function main() {
var i = 0, n2 = 0, n3 = 0, n5 = 0;
var u, ug2 = 1n, ug3 = 1n, ug5 = 1n;
var ugly = [];
for (;;) {
u = ug2;
if (u > ug3) u = ug3;
if (u > ug5) u = ug5;
ugly[i++] = u;
std.printf("%d: %s\n", i, String(u));
if (u >= 0x100000000000000000000000000000000n)
break;
if (u == ug2)
ug2 = 2n * ugly[n2++];
if (u == ug3)
ug3 = 3n * ugly[n3++];
if (u == ug5)
ug5 = 5n * ugly[n5++];
}
return 0;
}
main();
here is my code , the idea is to divide the number by 2 (till it gives remainder 0) then 3 and 5 . If at last the number becomes one it's a ugly number.
you can count and even print all ugly numbers till n.
int count = 0;
for (int i = 2; i <= n; i++) {
int temp = i;
while (temp % 2 == 0) temp=temp / 2;
while (temp % 3 == 0) temp=temp / 3;
while (temp % 5 == 0) temp=temp / 5;
if (temp == 1) {
cout << i << endl;
count++;
}
}
This problem can be done in O(1).
If we remove 1 and look at numbers between 2 through 30, we will notice that there are 22 numbers.
Now, for any number x in the 22 numbers above, there will be a number x + 30 in between 31 and 60 that is also ugly. Thus, we can find at least 22 numbers between 31 and 60. Now for every ugly number between 31 and 60, we can write it as s + 30. So s will be ugly too, since s + 30 is divisible by 2, 3, or 5. Thus, there will be exactly 22 numbers between 31 and 60. This logic can be repeated for every block of 30 numbers after that.
Thus, there will be 23 numbers in the first 30 numbers, and 22 for every 30 after that. That is, first 23 uglies will occur between 1 and 30, 45 uglies will occur between 1 and 60, 67 uglies will occur between 1 and 30 etc.
Now, if I am given n, say 137, I can see that 137/22 = 6.22. The answer will lie between 6*30 and 7*30 or between 180 and 210. By 180, I will have 6*22 + 1 = 133rd ugly number at 180. I will have 154th ugly number at 210. So I am looking for 4th ugly number (since 137 = 133 + 4)in the interval [2, 30], which is 5. The 137th ugly number is then 180 + 5 = 185.
Another example: if I want the 1500th ugly number, I count 1500/22 = 68 blocks. Thus, I will have 22*68 + 1 = 1497th ugly at 30*68 = 2040. The next three uglies in the [2, 30] block are 2, 3, and 4. So our required ugly is at 2040 + 4 = 2044.
The point it that I can simply build a list of ugly numbers between [2, 30] and simply find the answer by doing look ups in O(1).
Here is another O(n) approach (Python solution) based on the idea of merging three sorted lists. The challenge is to find the next ugly number in increasing order. For example, we know the first seven ugly numbers are [1,2,3,4,5,6,8]. The ugly numbers are actually from the following three lists:
list 1: 1*2, 2*2, 3*2, 4*2, 5*2, 6*2, 8*2 ... ( multiply each ugly number by 2 )
list 2: 1*3, 2*3, 3*3, 4*3, 5*3, 6*3, 8*3 ... ( multiply each ugly number by 3 )
list 3: 1*5, 2*5, 3*5, 4*5, 5*5, 6*5, 8*5 ... ( multiply each ugly number by 5 )
So the nth ugly number is the nth number of the list merged from the three lists above:
1, 1*2, 1*3, 2*2, 1*5, 2*3 ...
def nthuglynumber(n):
p2, p3, p5 = 0,0,0
uglynumber = [1]
while len(uglynumber) < n:
ugly2, ugly3, ugly5 = uglynumber[p2]*2, uglynumber[p3]*3, uglynumber[p5]*5
next = min(ugly2, ugly3, ugly5)
if next == ugly2: p2 += 1 # multiply each number
if next == ugly3: p3 += 1 # only once by each
if next == ugly5: p5 += 1 # of the three factors
uglynumber += [next]
return uglynumber[-1]
STEP I: computing three next possible ugly numbers from the three lists
ugly2, ugly3, ugly5 = uglynumber[p2]*2, uglynumber[p3]*3, uglynumber[p5]*5
STEP II, find the one next ugly number as the smallest of the three above:
next = min(ugly2, ugly3, ugly5)
STEP III: moving the pointer forward if its ugly number was the next ugly number
if next == ugly2: p2+=1
if next == ugly3: p3+=1
if next == ugly5: p5+=1
note: not using if with elif nor else
STEP IV: adding the next ugly number into the merged list uglynumber
uglynumber += [next]

Resources