I am looking at an application of catalan numbers:
Number of ways to form a “mountain ranges” with n upstrokes and n down-strokes that all stay above the original line.
Now given a number n, find the number of mountain ranges.
public int countMountainRanges(int n) {
}
What logic or formula can we use here to get the number of ways for input n.
I tried the formula F(n) = F(n-1) + F(n-2), but it does not work in this case.
F(n) = F(n-1) + F(n-2) is the formula for the nth Fibonacci number. The nth Catalan number, on the other hand, is given by (2n choose n) / (n + 1).
public static int countMountainRanges(int n) {
return choose(2 * n , n) / (n + 1);
}
private static int choose(int n, int k){
int res = 1;
k = Math.min(k, n - k);
for (int i = 0; i < k; i++) {
res = res * (n - i) / (i + 1);
}
return res;
}
I'm working on a problem where I'm required to manipulate large lists of palindromes up to a certain number of digits. This should work with numbers up 15 digits. The most common method I've seen for this is iterating through each number and checking whether each is a palindrome and then adding that to a list. This is my implementation in java and it works fine.
public class Palindrome {
public ArrayList<Long> List = new ArrayList<Long>();
public double d;
public Palindrome(double digits) {
this.d = digits;
long dig = (int)(Math.pow(10,digits));
for (long i = 1; i <= dig; i++) {
long a = i;
long b = inverse(a);
if (a == b) {
List.add(a);
}
}
public long inverse(long x){
long inv = 0;
while (x > 0) {
inv = inv * 10 + x % 10;
x = x / 10;
}
return inv;
}
}
Only problem is it's pretty slow when I get to 10+ digit palindromes. I've been considering alternative ways to create this list and one consideration I've had is generating the list of palindromes rather than iterating through each number and checking if it's a palindrome.
I'm still working on paper but the pattern isn't as obvious as I thought I would find it to turn into pseudocode. I'm working it out that for n number of digits, going from i to n, if the number of digits is even, generate numbers from 1 up to [10^(i/2 + 1) - 1]. Then append the reverse of each number to itself. A little stuck on how to do it for the odd digits. That's where I am right now.
I will come back with my own response if I figure this out and implement the code but in the meantime, I would just like to know if anyone has done this before or has an alternative method I've overlooked that would be more efficient.
UPDATE
So I did manage to work out something thanks to all your suggestions. I decided to work with the numbers as strings but contrary to what I intended this has actually increased the runtime :/
public class Palindrome2 {
public ArrayList<Long> List = new ArrayList<Long>();
public double d;
public Palindrome2(double digits) {
this.d = digits;
for (long n = 1; n <= d; n++) {
if (n == 1) {
for (long i = 1; i < 10; i++) {
List.add(i);
}
}
if (n % 2 != 0 && n != 1) {
long max = (long) Math.pow(10, (n + 1) / 2);
long min = (long) Math.pow(10, Math.floor(n / 2));
for (long i = min; i < max; i++) {
String str = Long.toString(i);
str = str + removeFirst(reverse(str));
Long x = Long.parseLong(str);
List.add(x);
}
} else if (n % 2 == 0) {
long max = (long) (Math.pow(10, Math.floor((n + 1) / 2)) - 1);
long min = (long) Math.pow(10, (n / 2) - 1);
for (long i = min; i <= max; i++) {
String str = Long.toString(i);
str = str + reverse(str);
Long x = Long.parseLong(str);
List.add(x);
}
}
}
}
public String reverse(String x) {
String rev = new StringBuffer(x).reverse().toString();
return rev;
}
public String removeFirst(String x) {
return x.substring(1);
}
}
Once again, accurate but still slow :(
Introduction
You need to analyzing the regular pattern for an algorithm roughly before jump into developing, that will saving lot of time, for example:
each 1 digit is 1 palindrome, e.g: 1
each 2 digits has 1 palindrome, e.g: 11.
each 3 digits has 10 palindromes, e.g: 101,111,...,191.
each 4 digits has 10 palindromes, e.g: 1001, 1111, ..., 1991.
each 5 digits has 100 palindromes, e.g: 10001, 11011, ..., 19091, ..., 19991.
each 6 digits has 100 palindromes, e.g: 100001, 110011, ..., 190091, ..., 199991.
each 7 digits has 1000 palindromes, e.g: 1000001, ...,1900091,...,1090901, ..., 1999991.
each 8 digits has 1000 palindromes, e.g: 10000001, ...,19000091,...,10900901, ..., 19999991.
....
then you can write some arrangement algorithm to implement this .
Implementation
But I can tell you this implementation can optimizing as further, if you using a cache to saving palindromes generated from low digits palindromes(2), then any high digits palindromes(n>2) can reusing it.
Maybe it's not robust but it pass all my tests on github. I left the rest working & optimization to you, and I wish you can done by yourself.
private static List<Integer> palindromes(int digits) {
return palindromes(digits, 0);
}
private static List<Integer> palindromes(int digits, int shifts) {
List<Integer> result = new ArrayList<>();
int radix = (int) Math.pow(10, digits - 1);
int renaming = digits - 2;
boolean hasRenaming = renaming > 0;
for (int i = start(digits, shifts); i <= 9; i++) {
int high = i * radix;
int low = low(digits, i);
if (hasRenaming) {
for (Integer m : palindromes(renaming, shifts + 1)) {
int ret = high + m * 10 + low;
if (ret < 0) {
return result;
}
result.add(ret);
}
} else {
result.add(high + low);
}
}
return result;
}
private static int low(int digits, int high) {
return digits > 1 ? high : 0;
}
private static int start(int digits, int shifts) {
return digits > 1 && shifts == 0 ? 1 : 0;
}
Usage
then you can collect all palindrome numbers as below:
// v--- min:0, max: 2147447412, count: 121474
List<Integer> all = IntStream.rangeClosed(1, 10)
.mapToObj(PalindromeTest::palindromes)
.flatMap(List::stream)
.collect(Collectors.toList());
Time Cost:
191ms
Enable Caching
public class Palindromes {
private static final int[] startingNonZerosTable = {
0,// 0
0, 1,// 1 2
10, 10,//3 4
100, 100, //5 6
1000, 1000,//7 8
10000, 10000,// 9 10
100000, 100000,//11 12
1000000, 1000000,//13 14
10000000, 10000000,//15 16
100000000, 100000000,//17 18
1000000000, 1000000000//19 20
};
private static final int MAX_DIGIT = 9;
private static final int MIN_DIGIT = 0;
private static final int RADIX = MAX_DIGIT - MIN_DIGIT + 1;
private static final int LONG_MAX_DIGITS = 19;
private static volatile long[][] cache = new long[LONG_MAX_DIGITS + 1][];
// includes palindromes(0) ---^
static {
cache[0] = new long[0];
cache[1] = new long[]{0L, 1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L};
cache[2] = new long[]{0L, 11L, 22L, 33L, 44L, 55L, 66L, 77L, 88L, 99L};
}
public static LongStream since1(int end) {
return between(1, end);
}
public static LongStream between(int start, int end) {
return IntStream.rangeClosed(start, end)
.mapToObj(Palindromes::of)
.flatMapToLong(identity());
}
public static LongStream of(int digits) {
return Arrays.stream(palindromes0(digits))
.skip(startingNonZerosTable[digits]);
}
private final static long[] palindromes0(int digits) {
if (cache[digits] != null) {
return cache[digits];
}
long[] result = new long[sizeOf(digits)];
int size = 0;
long high = (long) Math.pow(RADIX, digits - 1);
for (int i = MIN_DIGIT; i <= MAX_DIGIT; i++) {
for (long mid : palindromes0(digits - 2)) {
long value = i * high + mid * RADIX + i;
if (value < 0) {//overflow
return cache[digits] = Arrays.copyOf(result, size);
}
result[size++] = value;
}
}
return cache[digits] = result;
}
private static int sizeOf(int digits) {
return MAX_DIGIT * (int) Math.pow(RADIX, (digits - 1) >>> 1)
+ startingNonZerosTable[digits];
}
// v--- java -Xms1024m -Xmx2048m test.algorithm.Palindromes
public static void main(String[] args) {
Duration duration = timing(() -> {
// palindromes[1..15] ---v
LongSummaryStatistics result = since1(15).summaryStatistics();
long max = result.getMax();
long count = result.getCount();
System.out.printf("Max: %d, Count: %d%n", max, count);
});
System.out.printf("Time Elapsed:%s%n", duration);
// ^--- time elapsed: 4s
}
private static Duration timing(Runnable task) {
long starts = System.currentTimeMillis();
task.run();
return Duration.ofMillis(System.currentTimeMillis() - starts);
}
}
Time Cost:
palindromes[1..15] time elapsed: 4s
Have you tried working with characters rather than numbers? You could generate the palindrome as a string of digits and then convert to a number at the end. Something like this pseudocode:
generatePalindrome(size)
half <- size DIV 2 // Integer division
result <- ""
result.append(randomDigitIn(1..9)) // No leading zeros.
while (result.length <= half)
result.append(randomDigitIn(0..9))
endwhile
if (size is odd)
result <- result + randomDigitIn(0..9) + result.reverse()
else
result <- result + result.reverse()
endif
return number.parse(result)
end generatePalindrome()
Basically you randomly generate half the palindrome, avoiding leading zeros, insert an extra digit in the middle for odd lengths, append the reversed first half and then parse the digit string into the number format you want.
For odd digit you can simply reuse the palindromes generated at the previous even step , split them in half and insert in the middle all the possible number from 0 to 9.
Let's say you need to generate the palindrom of 3 digit, simply get all the palindromes of 2 digit and add insert all the number from 0 to 9.
We have 22 than we can generate:
202
212
222
232
and so on
Hope my idea is clear:)
Try something like this:
public class Palindrome
{
public static ArrayList<Long> calculatePalindromes(int maxLength) {
ArrayList<Long> result = new ArrayList<>();
if (maxLength <= 0) {
return result;
}
long maxPart = (long)Math.pow(10, maxLength / 2);
for (long i = 0; i < 10; ++i) {
result.add(i);
}
for (long i = 1; i < maxPart; ++i) {
long curHalf = i;
long curNum = i;
int curLen = 0;
while (curHalf != 0) {
curNum *= 10;
curNum += curHalf % 10;
curHalf /= 10;
++curLen;
}
result.add(curNum);
// insert numbers from 0 to 9
if (curLen * 2 + 1 > maxLength) {
continue;
}
for (int j = 0; j < 10; ++j) {
curHalf = i;
curNum = i;
curNum *= 10;
curNum += j;
while (curHalf != 0) {
curNum *= 10;
curNum += curHalf % 10;
curHalf /= 10;
}
result.add(curNum);
}
}
return result;
}
}
The idea is to insert numbers from 0 to 9 after each X and add reversed(X) after it so we get X (1..9) reversed(X).
You can generate all palindromes in the needed range without check, but you will probably face with the memory insufficiency, as storing all these numbesr for 15-length upper number in the list - is a bad idea.
More specifically your code will looks like:
long dig = (long) Math.pow(10, digits / 2);
int pow = 10;
int npow = 100;
for (long i = 1; i <= dig; i++) {
System.out.println(i * pow + inverse(i));
System.out.println(i * pow / 10 + inverse(i / 10));
// list.add(i * pow + inverse(i));
// list.add(i * pow/10 + inverse(i / 10));
if (i % pow == 0) {
pow = npow;
npow *= 10;
}
}
I have deliberately commented list adding lines.
The idea is to push into list/output all numbers composed with given half as:
XXXY+YXXX
and
XXX+Y+XXX
i.e. generating both cases: odd and even palindromes.
I am writing code for counting the number of ways an integer can be represented as a sum of the consecutive integers. For Example
15=(7+8),(1+2+3+4+5),(4+5+6). So the number of ways equals 3 for 15.
Now the input size can be <=10^12. My program is working fine till 10^7(i think so, but not sure as i didnt check it on any online judge. Feel free to check the code for that)
but as soon as the i give it 10^8 or higher integer as input. it throws many runtime exceptions(it doesnt show what runtime error). Thanks in advance.
import java.io.*;
//sum needs to contain atleast 2 elements
public class IntegerRepresentedAsSumOfConsecutivePositiveIntegers
{
public static long count = 0;
public static void main(String[] args) throws IOException
{
BufferedReader br = new BufferedReader(new InputStreamReader(System.in));
long num = Long.parseLong(br.readLine()); //Enter a number( <=10^12)
driver(num);
System.out.println("count = " + count);
}
public static void driver(long num)
{
long limit = num / 2;
for(long i = 1 ; i <= limit ; i++)
{
func(i,num);
}
}
public static void func(long i,long num)
{
if(i < num)
{
func(i + 1,num - i);
}
else if(i > num)
{
return;
}
else
{
count++;
}
}
}
Use some math: if arithmetic progression with difference 1 starts with a0 and contains n items, then its sum is
S = (2 * a0 + (n-1))/2 * n = a0 * n + n * (n-1) / 2
note that the second summand rises as quadratic function. So instead of checking all a0 in range S/2, we can check all n is smaller range
nmax = Ceil((-1 + Sqrt(1 + 8 * S)) / 2)
(I used some higher approximation).
Just test whether next expression gives integer positive result
a0 = (S - n * (n - 1) / 2) / n
Recursive function isn't suitable when you have big input size like your case.
The maximum depth of the java call stack is about 8900 calls and sometimes only after 7700 calls stack overflow occurs so it really depends on your program input size.
Try this algorithm I think it worked for your problem:
it will work fine until 10^9 after that it will take much more time to finish running the program.
long sum = 0;
int count = 0;
long size;
Scanner in = new Scanner(System.in);
System.out.print("Enter a number <=10^12: ");
long n = in.nextLong();
if(n % 2 != 0){
size = n / 2 + 1;
}
else{
size = n / 2;
}
for(int i = 1; i <= size; i++){
for(int j = i; j <= size; j++){
sum = sum + j;
if(sum == n){
sum = 0;
count++;
break;
}
else if(sum > n){
sum = 0;
break;
}
}
}
System.out.println(count);
Output:
Enter a number <=10^12: 15
3
Enter a number <=10^12: 1000000000
9
BUILD SUCCESSFUL (total time: 10 seconds)
There's a really excellent proof that the answer can be determined by solving for the unique odd factors (Reference). Essentially, for every odd factor of a target value, there exists either an odd series of numbers of that factor multiplied by its average to produce the target value, or an odd average equal to that factor that can be multiplied by double an even-sized series to reach the target value.
public static int countUniqueOddFactors(long n) {
if (n==1) return 1;
Map<Long, Integer> countFactors=new HashMap<>();
while ((n&1)==0) n>>>=1; // Eliminate even factors
long divisor=3;
long max=(long) Math.sqrt(n);
while (divisor <= max) {
if (n % divisor==0) {
if (countFactors.containsKey(divisor)) {
countFactors.put(divisor, countFactors.get(divisor)+1);
} else {
countFactors.put(divisor, 1);
}
n /= divisor;
} else {
divisor+=2;
}
}
int factors=1;
for (Integer factorCt : countFactors.values()) {
factors*=(factorCt+1);
}
return factors;
}
As #MBo noted, if a number S can be partitioned into n consecutive parts, then S - T(n) must be divisible by n, where T(n) is the n'th triangular number, and so you can count the number of partitions in O(sqrt(S)) time.
// number of integer partitions into (at least 2) consecutive parts
static int numberOfTrapezoidalPartitions(final long sum) {
assert sum > 0: sum;
int n = 2;
int numberOfPartitions = 0;
long triangularNumber = n * (n + 1) / 2;
while (sum - triangularNumber >= 0) {
long difference = sum - triangularNumber;
if (difference == 0 || difference % n == 0)
numberOfPartitions++;
n++;
triangularNumber += n;
}
return numberOfPartitions;
}
A bit more math yields an even simpler way. Wikipedia says:
The politeness of a positive number is defined as the number of ways it can be expressed as the sum of consecutive integers. For every x, the politeness of x equals the number of odd divisors of x that are greater than one.
Also see: OEIS A069283
So a simple solution with lots of room for optimization is:
// number of odd divisors greater than one
static int politeness(long x) {
assert x > 0: x;
int p = 0;
for (int d = 3; d <= x; d += 2)
if (x % d == 0)
p++;
return p;
}
I'm new with algorithms and wonder how this displayed function is supposed to be converted/transformed from recursive to iterative. What should I keep in mind when converting?
public int binom(int n, int k)
{
if (k == 0 || n == k) { return 1; }
return binom(n - 1, k - 1) + binom(n - 1, k);
}
Thanks in advance!
In fact, this problem is not so easy, if you just look at the recursive code and try to decrypt it.
However, it might be a helpful hint for you, that (n over k), i.e. the binomial coefficient can be written as
n! / (k! * (n - k)!)
where "!" denotes the factorial.
And it should be rather easy to compute the factorial in a Loop (i.e. iterative) for you.
If intermediate results are too big you can shorten before computation. You can shorten either term k! or the term (n-k)! (you would choose the bigger one). For example with n = 5 and k = 3 you have:
(1 * 2 * 3 * 4 * 5) / ((1 * 2 * 3) * (1 * 2)) = (4 * 5) / (1 * 2)
Spoiler-Alarm:
public static int binomial(int n, int k) {
int nMinusK = n - k;
if (n < nMinusK) {
//Switch n and nMinusK
int temp = n;
n = nMinusK;
nMinusK = temp;
}
int result = 1;
// n!/k!
for (int i = k + 1; i <= n; i++) {
result *= i;
}
//Division by (n-k)!
for (int j = 1; j <= nMinusK; j++) {
result = result / j;
}
return result;
}
You can use the multiplicative form of binomial coefficients, for example from Wikia, which can be easily implemented with faculties or loops.
I am trying to find the Largest prime factor of a number while solving this problem here. I think that I am doing everything right, however one of the test case (#2) is failing and I can't think of any corner case where it might fail. Here's my code, please have a look and try to spot something.
public class ProblemThree
{
public static void main(String[] args)
{
Scanner scanner = new Scanner(System.in);
int T = scanner.nextInt();
for (int i = 0; i < T; i++)
{
System.out.println(largestPrime(scanner.nextLong()));
}
}
private static long largestPrime(long n)
{
while (n % 2 == 0)
{
n = n / 2; // remove all the multiples of 2
}
while (n % 3 == 0)
{
n = n / 3; // remove all the multiples of 2
}
// remove multiples of prime numbers other than 2 and 3
while (n >= 5)
{
boolean isDivisionComplete = true;
for (long i = 5; i < Math.ceil(Math.sqrt(n)); i++)
{
if (n % i == 0)
{
n = n / i;
isDivisionComplete = false;
break;
}
}
if (isDivisionComplete)
{
break;
}
}
return n;
}
}
Basically, what I am doing is:
Largest_Prime(n):
1. Repeatedly divide the no by any small number, say x where 0 < x < sqrt(n).
2. Then set n = n/x and repeat steps 1 and 2 until there is no such x that divides n.
3 Return n.
It seems you have some bug in your code as as when you input 16 largestPrime function return 1. and this is true for when input is the power of 3.
Detailed Algorithm description:
You can do this by keeping three variables:
The number you are trying to factor (A)
A current divisor store (B)
A largest divisor store (C)
Initially, let (A) be the number you are interested in - in this case, it is 600851475143. Then let (B) be 2. Have a conditional that checks if (A) is divisible by (B). If it is divisible, divide (A) by (B), reset (B) to 2, and go back to checking if (A) is divisible by (B). Else, if (A) is not divisible by (B), increment (B) by +1 and then check if (A) is divisible by (B). Run the loop until (A) is 1. The (3) you return will be the largest prime divisor of 600851475143.
public static void main(String[] args) {
Scanner in = new Scanner(System.in);
int t = in.nextInt();
for(int a0 = 0; a0 < t; a0++){
long n = in.nextLong();
long A=n;
long B=2;
long C=0;
while(Math.pow(B,2)<=A)
{
if(A%B==0)
{
C=B;
A=A/B;
B=2;
}
else
B++;
}
if(A>=C)
C=A;
if(A==1)
{ C=2;
break;
}
System.out.println(C);
}
}
Why are you removing multiples of 2 and multiples of 3? This way if you have a number that is any combination of powers of 2 and 3 you will get your answer as 1 which is clearly wrong.
For this problem you can do the naive way of looping from 2 to sqrt(n) and store the largest number which divides n, when you finish your loop just return the highest divisor you found.
1 drop your loop for 2 and 3. If not, you dont get 2, 2x2, 3, 2x3, ... all multiples of 2 and 3
2 change your loop to stop at 2 (and not 5):
while (n >= 2)
{
3 stop if 2
if (n==2) return 2;
4 loop from 2
and
5 loop until sqrt(n), with <= and not only < (if not, you dont get prime X Prime)
for (long i = 2; i <= Math.ceil(Math.sqrt(n)); i++)
One easy way of extracting prime factors is like this:
/**
* Prime factors of the number - not the most efficient but it works.
*
* #param n - The number to factorise.
* #param unique - Want only unique factors.
* #return - List of all prime factors of n.
*/
public static List<Long> primeFactors(long n, boolean unique) {
Collection<Long> factors;
if (unique) {
factors = new HashSet<>();
} else {
factors = new ArrayList<>();
}
for (long i = 2; i <= n / i; i++) {
while (n % i == 0) {
factors.add(i);
n /= i;
}
}
if (n > 1) {
factors.add(n);
}
return new ArrayList<>(factors);
}
Those first loops are a problem. They will reduce all even numbers to 1 - thus missing 2 as the factor. Changing your code to use:
while (n > 2 && n % 2 == 0) {
n = n / 2; // remove all the multiples of 2
}
while (n > 3 && n % 3 == 0) {
n = n / 3; // remove all the multiples of 2
}
You still have further issues - e.g. you report the largest prime factor of 25 to be 25 and the largest prime factor of 49 to be 49.
Just run this code using yours and mine to see where yours fails:
for (long i = 1; i < 1000; i++) {
long largestPrime = largestPrime(i);
List<Long> primeFactors = primeFactors(i, true);
if (primeFactors.size() > 0) {
Collections.sort(primeFactors, Collections.reverseOrder());
long highestFactor = primeFactors.get(0);
if (largestPrime != highestFactor) {
System.out.println("Wrong! " + i + " " + largestPrime + " != " + primeFactors);
}
} else {
System.out.println("No factors for " + i);
}
}