Open In App
Related Articles

Median of two Sorted Arrays of Different Sizes using Binary Search

Improve
Improve
Improve
Like Article
Like
Save Article
Save
Report issue
Report

Given two sorted arrays, a[] and b[] os size n and m, the task is to find the median of these sorted arrays, in O(log(min(n, m)), when n is the number of elements in the first array, and m is the number of elements in the second array.

Note: In case of even numbers in total and if we want to return a median that exists in the merged array we can return the element in the (n+m)/2 or ((n+m)/2 – 1)th position.

Examples:

Input: a[] = {-5, 3, 6, 12, 15}, b[] = {-12, -10, -6, -3, 4, 10}
Output : The median is 3.
Explanation: The merged array is {-12, -10, -6, -5 , -3, 3, 4, 6, 10, 12, 15}, with median 3

Input : a[] = {2, 3, 5, 8}, b[] = {10, 12, 14, 16, 18, 20}
Output : The median is 11.
Explanation: The merged array is {2, 3, 5, 8, 10, 12, 14, 16, 18, 20}
Median = average of two middle elements (as number of the elements are even) = (10 + 12) / 2 = 11.

Binary Search Approach to find Median of Two Sorted Arrays of Different Sizes:

Let us consider the case for a total of odd number of elements. Also consider, that i elements from the first array and j elements from the 2nd array are on the left half of the median (including median itself). Then the maximum of a[i] and b[j] is the median, The important task is to efficiently find out such i and j.

How to efficiently find i and j? 

Let us consider there are i elements from the first array that are on the left of the median and the other elements on the left are from the second array. So the number of other elements are j = (n + m + 1)/2 – i. Now we need to validate if our assumption is correct. For that we have take the following actions:

  • If a[i+1] > b[j] and b[j+1] > a[i] then we can definitely say that these are the elements on the left of the median.
  • Otherwise, we have to reduce the number of elements taken from any one array. That can be decided based on the following fact:
    • If a[i+1] is less than b[j], then we need to consider more elements from the first array.
    • Otherwise, if b[j+1] is less than a[i], we need to consider more elements from the second array.

Instead of linear incrementation for considering more elements from any of the array, we can use Binary Search here.

Consider, in a previous search we have found x1 elements must be considered from the first array, and we consider not more than x2 elements can be chose from the first array. Now we are checking for i = (x1 + x2)/2 elements from the first array.

  • If a[i+1] < b[j], we will change x1 = i+1.
  • If the reverse is true, i.e., b[j+1] < a[i] make x2 = i-1.

So the solution comes down to the following:

Initially take x1 = 0 and x2 = N and repeatedly follow the above binary search concept repeatedly, till we find out such i and j. After that the maximum between a[i] and b[j] will be the median when there are total odd number of elements.

Note that in case the total number of elements is even, the median will be the average of max(a[i], b[j]) and min(a[i+1], b[j+1]).

Note: The partition didn’t work if any one array is empty from the given arrays:

For example: if arr1=[2], arr2=[] by this  “(n + m + 1) / 2” formula the value of i=0 and value of j=1 and this give you out of index error because arr2 is empty and arr2[j] give you out of index error. You have to handle this case by checking if one array is empty you can simply return the medium of another array.

What to be careful about during calculation?

There is one scope of error while calculating j = (n + m + 1)/2 – i. This may result in a negative value of j for some cases. So to avoid accessing a negative index, check the value of j, and if it is negative, the median is on the first array itself. Find the element at the position (n + m + 1)/2 [the average of (n+m)/2 and the next element if total even number of elements] and return this as the median.

Illustration of finding Median of two sorted arrays of different size using Binary Search

See the below examples for a better understanding:

Example 1: When the total number of elements is odd:

Step 1: Initially, consider that 3 elements from the first array are to the left of the median. But this does not fit the condition as 20 > 11. So, for the next iteration, a minimum of 4 elements from the first array should be considered.

Consider 3 elements from 1st array on the left of the median

Consider 3 elements from 1st array on the left of the median

Step 2: A minimum of 4 and not more than 5 from the first array can be considered. So we consider 4 elements from the first array. This satisfies the condition and both 11 and 13 are less than 20 and 17 respectively. So the median is max(11, 13) = 13.

Consider 4 elements from 1st array on the left of the median

Consider 4 elements from 1st array on the left of the median

Example 2: When the total number of elements is even.

Step 1: Consider 2 elements from the first array to be on the left of the median. But 14 > 5. So more elements from the first array should be considered.

Consider 2 elements from the first on the left of the median

Consider 2 elements from the first on the left of the median

Step 2: At least 3 and not more than 4 elements from the first array can be considered. So, we consider 3 elements. But again 12 > 8. So we need to take more from the first array.

Consider 3 elements from the first on the left of the median

Consider 3 elements from the first on the left of the median

Step 3: We need to consider more than 3 from the first array. So we consider all the elements from the first array and this satisfies the condition that 8 < 12. So now the median becomes the average of max(8, 10) and 12 i.e., (10 + 12)/2 = 11.

Consider all elements from the first on the left of the median

Consider all elements from the first on the left of the median

Code implementation of the above approach:

C++

// CPP code for median with case of returning
// double value when even number of elements are
// present in both array combinely
#include <bits/stdc++.h>
using std::cout;
 
int maximum(int a, int b);
int minimum(int a, int b);
 
// Function to find median of two sorted arrays
double findMedianSortedArrays(int* a, int n, int* b, int m)
{
 
    int min_index = 0, max_index = n, i, j, median;
 
    while (min_index <= max_index) {
        i = (min_index + max_index) / 2;
        j = ((n + m + 1) / 2) - i;
 
        // If j is negative then the partition is not
        // possible having i elements from array i
        if (j < 0) {
            max_index = i - 1;
            continue;
        }
 
        // if i = n, it means that Elements from a[] in
        // the second half is an empty set. and if j = 0,
        // it means that Elements from b[] in the first
        // half is an empty set. so it is necessary to
        // check that, because we compare elements from
        // these two groups.
        // Searching on right
        if (i < n && j > 0 && b[j - 1] > a[i])
            min_index = i + 1;
 
        // if i = 0, it means that Elements from a[] in
        // the first half is an empty set and if j = m,
        // it means that Elements from b[] in the second
        // half is an empty set. so it is necessary to
        // check that, because we compare elements
        // from these two groups.
        // searching on left
        else if (i > 0 && j < m && b[j] < a[i - 1])
            max_index = i - 1;
 
        // we have found the desired halves.
        else {
            // this condition happens when we don't have any
            // elements in the first half from a[] so we
            // returning the last element in b[] from
            // the first half.
            if (i == 0)
                median = b[j - 1];
 
            // and this condition happens when we don't
            // have any elements in the first half from
            // b[] so we returning the last element in
            // a[] from the first half.
            else if (j == 0)
                median = a[i - 1];
            else
                median = maximum(a[i - 1], b[j - 1]);
            break;
        }
    }
 
    // calculating the median.
    // If number of elements is odd there is
    // one middle element.
    if ((n + m) % 2 == 1)
        return (double)median;
 
    // Elements from a[] in the second half is an empty set.
    if (i == n)
        return (median + b[j]) / 2.0;
 
    // Elements from b[] in the second half is an empty set.
    if (j == m)
        return (median + a[i]) / 2.0;
 
    return (median + minimum(a[i], b[j])) / 2.0;
}
 
// Function to find max
int maximum(int a, int b) { return a > b ? a : b; }
 
// Function to find minimum
int minimum(int a, int b) { return a < b ? a : b; }
 
// Driver code
int main()
{
    int a[] = { 900 };
    int b[] = { 10, 13, 14 };
    int n = sizeof(a) / sizeof(int);
    int m = sizeof(b) / sizeof(int);
 
    // we need to define the smaller array as the
    // first parameter to make sure that the
    // time complexity will be O(log(min(n,m)))
    if (n < m)
        cout << "The median is : "
             << findMedianSortedArrays(a, n, b, m);
    else
        cout << "The median is : "
             << findMedianSortedArrays(b, m, a, n);
 
    return 0;
}

                    

Java

// Java code for median with
// case of returning double
// value when even number of
// elements are present in
// both array combinely
import java.io.*;
 
class GFG {
    static int[] a = new int[] { 900 };
    static int[] b = new int[] { 10, 13, 14 };
 
    // Function to find max
    static int maximum(int a, int b)
    {
        return a > b ? a : b;
    }
 
    // Function to find minimum
    static int minimum(int a, int b)
    {
        return a < b ? a : b;
    }
 
    // Function to find median
    // of two sorted arrays
    static double findMedianSortedArrays(int n, int m)
    {
 
        int min_index = 0, max_index = n, i = 0, j = 0,
            median = 0;
 
        while (min_index <= max_index) {
            i = (min_index + max_index) / 2;
            j = ((n + m + 1) / 2) - i;
 
            // if i = n, it means that Elements
            // from a[] in the second half is an
            // empty set. and if j = 0, it means
            // that Elements from b[] in the first
            // half is an empty set. so it is
            // necessary to check that, because we
            // compare elements from these two
            // groups. Searching on right
            if (i < n && j > 0 && b[j - 1] > a[i])
                min_index = i + 1;
 
            // if i = 0, it means that Elements
            // from a[] in the first half is an
            // empty set and if j = m, it means
            // that Elements from b[] in the second
            // half is an empty set. so it is
            // necessary to check that, because
            // we compare elements from these two
            // groups. searching on left
            else if (i > 0 && j < m && b[j] < a[i - 1])
                max_index = i - 1;
 
            // we have found the desired halves.
            else {
                // this condition happens when we
                // don't have any elements in the
                // first half from a[] so we
                // returning the last element in
                // b[] from the first half.
                if (i == 0)
                    median = b[j - 1];
 
                // and this condition happens when
                // we don't have any elements in the
                // first half from b[] so we
                // returning the last element in
                // a[] from the first half.
                else if (j == 0)
                    median = a[i - 1];
                else
                    median = maximum(a[i - 1], b[j - 1]);
                break;
            }
        }
 
        // calculating the median.
        // If number of elements is odd
        // there is one middle element.
        if ((n + m) % 2 == 1)
            return (double)median;
 
        // Elements from a[] in the
        // second half is an empty set.
        if (i == n)
            return (median + b[j]) / 2.0;
 
        // Elements from b[] in the
        // second half is an empty set.
        if (j == m)
            return (median + a[i]) / 2.0;
 
        return (median + minimum(a[i], b[j])) / 2.0;
    }
 
    // Driver code
    public static void main(String args[])
    {
        int n = a.length;
        int m = b.length;
 
        // we need to define the
        // smaller array as the
        // first parameter to
        // make sure that the
        // time complexity will
        // be O(log(min(n,m)))
        if (n < m)
            System.out.print(
                "The median is : "
                + findMedianSortedArrays(n, m));
        else
            System.out.print(
                "The median is : "
                + findMedianSortedArrays(m, n));
    }
}
 
// This code is contributed by
// Manish Shaw(manishshaw1)

                    

Python3

# Python code for median with
# case of returning double
# value when even number
# of elements are present
# in both array combinely
median = 0
i = 0
j = 0
 
# def to find max
 
 
def maximum(a, b):
    return a if a > b else b
 
# def to find minimum
 
 
def minimum(a, b):
    return a if a < b else b
 
# def to find median
# of two sorted arrays
 
 
def findMedianSortedArrays(a, n, b, m):
 
      # if a is empty
      if n == 0:
      # if b has even no. of elements
          if m % 2 == 0:
            return (b[m/2]+b[(m/2)+1])/2
      # if b has odd no. of elements
          else:
            return b[int(m/2)]
 
     # if b is empty
      elif m == 0:
      # if a has even no. of elements
           if n % 2 == 0:
              return (a[n/2]+a[(n/2)+1])/2
      # if a has odd no. of elements
      else:
           return a[int(n/2)]
 
 
global median, i, j
min_index = 0
max_index = n
 
while (min_index <= max_index):
 
        i = int((min_index + max_index) / 2)
        j = int(((n + m + 1) / 2) - i)
 
        # if i = n, it means that
        # Elements from a[] in the
        # second half is an empty
        # set. and if j = 0, it
        # means that Elements from
        # b[] in the first half is
        # an empty set. so it is
        # necessary to check that,
        # because we compare elements
        # from these two groups.
        # Searching on right
        if (i < n and j > 0 and b[j - 1] > a[i]):
            min_index = i + 1
 
        # if i = 0, it means that
        # Elements from a[] in the
        # first half is an empty
        # set and if j = m, it means
        # that Elements from b[] in
        # the second half is an empty
        # set. so it is necessary to
        # check that, because we compare
        # elements from these two groups.
        # searching on left
        elif (i > 0 and j < m and b[j] < a[i - 1]):
            max_index = i - 1
 
        # we have found the
        # desired halves.
        else:
 
            # this condition happens when
            # we don't have any elements
            # in the first half from a[]
            # so we returning the last
            # element in b[] from the
            # first half.
            if (i == 0):
                median = b[j - 1]
 
            # and this condition happens
            # when we don't have any
            # elements in the first half
            # from b[] so we returning the
            # last element in a[] from the
            # first half.
            elif (j == 0):
                median = a[i - 1]
            else:
                median = maximum(a[i - 1], b[j - 1])
            break
 
    # calculating the median.
    # If number of elements
    # is odd there is
    # one middle element.
 
    if ((n + m) % 2 == 1):
         return median
 
    # Elements from a[] in the
    # second half is an empty set.
    if (i == n):
        return ((median + b[j]) / 2.0)
 
    # Elements from b[] in the
    # second half is an empty set.
    if (j == m):
        return ((median + a[i]) / 2.0)
 
    return ((median + minimum(a[i], b[j])) / 2.0)
 
 
# Driver code
a = [900]
b = [10, 13, 14]
n = len(a)
m = len(b)
 
# we need to define the
# smaller array as the
# first parameter to make
# sure that the time complexity
# will be O(log(min(n,m)))
if (n < m):
    print("The median is : {}".format(findMedianSortedArrays(a, n, b, m)))
else:
    echo("The median is : {}".format(findMedianSortedArrays(b, m, a, n)))
 
# This code is contributed
# by Aditya Khare(adityaskhare123)

                    

C#

// C# code for median with case of returning
// double value when even number of elements
// are present in both array combinely
using System;
 
class GFG {
 
    // Function to find max
    static int maximum(int a, int b)
    {
        return a > b ? a : b;
    }
 
    // Function to find minimum
    static int minimum(int a, int b)
    {
        return a < b ? a : b;
    }
 
    // Function to find median of two sorted
    // arrays
    static double findMedianSortedArrays(ref int[] a, int n,
                                         ref int[] b, int m)
    {
 
        int min_index = 0, max_index = n, i = 0, j = 0,
            median = 0;
 
        while (min_index <= max_index) {
            i = (min_index + max_index) / 2;
            j = ((n + m + 1) / 2) - i;
 
            // if i = n, it means that Elements
            // from a[] in the second half is an
            // empty set. and if j = 0, it means
            // that Elements from b[] in the first
            // half is an empty set. so it is
            // necessary to check that, because we
            // compare elements from these two
            // groups. Searching on right
            if (i < n && j > 0 && b[j - 1] > a[i])
                min_index = i + 1;
 
            // if i = 0, it means that Elements
            // from a[] in the first half is an
            // empty set and if j = m, it means
            // that Elements from b[] in the second
            // half is an empty set. so it is
            // necessary to check that, because
            // we compare elements from these two
            // groups. searching on left
            else if (i > 0 && j < m && b[j] < a[i - 1])
                max_index = i - 1;
 
            // we have found the desired halves.
            else {
                // this condition happens when we
                // don't have any elements in the
                // first half from a[] so we
                // returning the last element in
                // b[] from the first half.
                if (i == 0)
                    median = b[j - 1];
 
                // and this condition happens when
                // we don't have any elements in the
                // first half from b[] so we
                // returning the last element in
                // a[] from the first half.
                else if (j == 0)
                    median = a[i - 1];
                else
                    median = maximum(a[i - 1], b[j - 1]);
                break;
            }
        }
 
        // calculating the median.
        // If number of elements is odd
        // there is one middle element.
        if ((n + m) % 2 == 1)
            return (double)median;
 
        // Elements from a[] in the second
        // half is an empty set.
        if (i == n)
            return (median + b[j]) / 2.0;
 
        // Elements from b[] in the second
        // half is an empty set.
        if (j == m)
            return (median + a[i]) / 2.0;
 
        return (median + minimum(a[i], b[j])) / 2.0;
    }
 
    // Driver code
    static void Main()
    {
        int[] a = new int[] { 900 };
        int[] b = new int[] { 10, 13, 14 };
        int n = a.Length;
        int m = b.Length;
 
        // we need to define the smaller
        // array as the first parameter to
        // make sure that the time
        // complexity will be O(log(min(n,m)))
        if (n < m)
            Console.Write("The median is : "
                          + findMedianSortedArrays(
                              ref a, n, ref b, m));
        else
            Console.Write("The median is : "
                          + findMedianSortedArrays(
                              ref b, m, ref a, n));
    }
}
 
// This code is contributed by Manish Shaw
// (manishshaw1)

                    

PHP

<?php
// PHP code for median with 
// case of returning double
// value when even number
// of elements are present
// in both array combinely
$median = 0;
$i = 0; $j = 0;
 
// Function to find max
function maximum($a, $b)
{
    return $a > $b ? $a : $b;
}
 
// Function to find minimum
function minimum($a, $b)
{
    return $a < $b ? $a : $b;
}
 
// Function to find median
// of two sorted arrays
function findMedianSortedArrays(&$a, $n,
                                &$b, $m)
{
    global $median, $i, $j;
    $min_index = 0;
    $max_index = $n;
     
    while ($min_index <= $max_index)
    {
        $i = intval(($min_index +
                     $max_index) / 2);
        $j = intval((($n + $m + 1) /
                           2) - $i);
     
        // if i = n, it means that
        // Elements from a[] in the
        // second half is an empty
        // set. and if j = 0, it
        // means that Elements from
        // b[] in the first half is
        // an empty set. so it is
        // necessary to check that,
        // because we compare elements
        // from these two groups.
        // Searching on right
        if ($i < $n && $j > 0 &&
            $b[$j - 1] > $a[$i])    
            $min_index = $i + 1;
                 
        // if i = 0, it means that
        // Elements from a[] in the
        // first half is an empty
        // set and if j = m, it means
        // that Elements from b[] in
        // the second half is an empty
        // set. so it is necessary to
        // check that, because we compare
        // elements from these two groups.
        // searching on left
        else if ($i > 0 && $j < $m &&
                 $b[$j] < $a[$i - 1])    
            $max_index = $i - 1;
         
        // we have found the
        // desired halves.
        else
        {
            // this condition happens when
            // we don't have any elements
            // in the first half from a[]
            // so we returning the last
            // element in b[] from the
            // first half.
            if ($i == 0)
                $median = $b[$j - 1];
                 
            // and this condition happens
            // when we don't have any
            // elements in the first half
            // from b[] so we returning the
            // last element in a[] from the
            // first half.
            else if ($j == 0)        
                $median = $a[$i - 1];        
            else       
                $median = maximum($a[$i - 1],
                                  $b[$j - 1]);
            break;
        }
    }
     
    // calculating the median.
    // If number of elements
    // is odd there is
    // one middle element.
 
    if (($n + $m) % 2 == 1)
        return $median;
 
    // Elements from a[] in the
    // second half is an empty set.
    if ($i == $n)
        return (($median +
                 $b[$j]) / 2.0);
 
    // Elements from b[] in the
    // second half is an empty set.
    if ($j == $m)
        return (($median +
                 $a[$i]) / 2.0);
     
    return (($median +
             minimum($a[$i],
                     $b[$j])) / 2.0);
}
 
// Driver code
$a = array(900);
$b = array(10, 13, 14);
$n = count($a);
$m = count($b);
 
// we need to define the
// smaller array as the
// first parameter to make
// sure that the time complexity
// will be O(log(min(n,m)))
if ($n < $m)
    echo ("The median is : " .
           findMedianSortedArrays($a, $n,
                                  $b, $m));
else
    echo ("The median is : " .
           findMedianSortedArrays($b, $m,
                                  $a, $n));
                                   
// This code is contributed
// by Manish Shaw(manishshaw1)
?>

                    

Javascript

<script>
// Javascript code for median with
// case of returning double
// value when even number of
// elements are present in
// both array combinely
     
    let a=[900];
    let b=[10, 13, 14];
     
     // Function to find max
    function maximum(a,b)
    {
        return a > b ? a : b;
    }
     
    // Function to find minimum
    function minimum(a,b)
    {
        return a < b ? a : b;
    }
     
    // Function to find median
    // of two sorted arrays
    function findMedianSortedArrays(n,m)
    {
        let min_index = 0,
            max_index = n, i = 0,
            j = 0, median = 0;
          
        while (min_index <= max_index)
        {
            i = Math.floor((min_index + max_index) / 2);
            j = Math.floor((n + m + 1) / 2) - i;
          
            // if i = n, it means that Elements
            // from a[] in the second half is an
            // empty set. and if j = 0, it means
            // that Elements from b[] in the first
            // half is an empty set. so it is
            // necessary to check that, because we
            // compare elements from these two
            // groups. Searching on right
            if (i < n && j > 0 && b[j - 1] > a[i])   
                min_index = i + 1;
                      
            // if i = 0, it means that Elements
            // from a[] in the first half is an
            // empty set and if j = m, it means
            // that Elements from b[] in the second
            // half is an empty set. so it is
            // necessary to check that, because
            // we compare elements from these two
            // groups. searching on left
            else if (i > 0 && j < m && b[j] < a[i - 1])   
                max_index = i - 1;
      
            // we have found the desired halves.
            else
            {
                // this condition happens when we
                // don't have any elements in the
                // first half from a[] so we
                // returning the last element in
                // b[] from the first half.
                if (i == 0)       
                    median = b[j - 1];       
                  
                // and this condition happens when
                // we don't have any elements in the
                // first half from b[] so we
                // returning the last element in
                // a[] from the first half.
                else if (j == 0)       
                    median = a[i - 1];       
                else  
                    median = maximum(a[i - 1],
                                     b[j - 1]);       
                break;
            }
        }
          
        // calculating the median.
        // If number of elements is odd
        // there is one middle element.
        if ((n + m) % 2 == 1)
            return median;
              
        // Elements from a[] in the
        // second half is an empty set.
        if (i == n)
            return (median + b[j]) / 2.0;
              
        // Elements from b[] in the
        // second half is an empty set.
        if (j == m)
            return (median + a[i]) / 2.0;
          
        return (median + minimum(a[i],
                                 b[j])) / 2.0;
    }
     
    // Driver code
    let n = a.length;
    let m = b.length;
     
    // we need to define the
    // smaller array as the
    // first parameter to
    // make sure that the
    // time complexity will
    // be O(log(min(n,m)))
    if (n < m)
        document.write("The median is : " +
                   findMedianSortedArrays(n, m));
    else
        document.write("The median is : " +
                   findMedianSortedArrays(m, n));
     
     
     
     
    // This code is contributed by unknown2108
</script>

                    

Output
The median is : 13.5

Time Complexity: O(log(min(n, m))
Auxiliary Space: O(1), the space complexity of this algorithm is O(1), as we are not using any extra space.



Last Updated : 12 Jun, 2023
Like Article
Save Article
Previous
Next
Share your thoughts in the comments
Similar Reads