Maximum Sum Subsequence of length k

Given an array sequence [A1, A2 …An], the task is to find the maximum possible sum of increasing subsequence S of length k such that S1<=S2<=S3………<=Sk.

Examples:

Input :
n = 8 k = 3
A=[8 5 9 10 5 6 21 8]
Output : 40
Possible Increasing subsequence of Length 3 with maximum possible sum is 9 10 21



Input :
n = 9 k = 4
A=[2 5 3 9 15 33 6 18 20]
Output : 62
Possible Increasing subsequence of Length 4 with maximum possible sum is 9 15 18 20

One thing that is clearly visible that it can be easily solved with dynamic programming and this problem is a simple variation of Longest Increasing Subsequence. If you are unknown of how to calculate the longest increasing subsequence then see the implementation going to the link.

Naive Approach:
In the brute force approach, first we will try to find all the subsequences of length k and will check whether they are increasing or not. There could be nCk such sequences in the worst case when all elements are in increasing order. Now we will find the maximum possible sum for such sequences.
Time Complexity would be O((nCk)*n).

Efficient Approach:
We will be using a two-dimensional dp array in which dp[i][l] means that maximum sum subsequence of length l taking array values from 0 to i and the subsequence is ending at index ‘i’. Range of ‘l’ is from 0 to k-1. Using the approach of longer increasing subsequence on the inner loop when j<i we will check if arr[j] < arr[i] for checking if subsequence increasing.
This problem can be divided into its subproblems:

dp[i][1]=arr[i] for length 1 , maximum icreasing subsequence is equal to the array value
dp[i][l+1]= max(dp[i][l+1], dp[j][l]+arr[i]) for any length l between 1 to k-1

This means that if for ith position and subsequence of length l+1 , there exists some subsequence at j (j < i) of length l for which sum of dp[j][l] + arr[i] is more than its initial calculated value then update that value.
Then finally we will find the maximum value of dp[i][k] i.e for every ‘i’ if subsequence of k length is causing more sum than update the required ans.

Below is the implementation code:

filter_none

edit
close

play_arrow

link
brightness_4
code

/*C++ program to calculate the maximum sum of 
increasing subsequence of length k*/
#include <bits/stdc++.h>
using namespace std;
int MaxIncreasingSub(int arr[], int n, int k)
{
    // In the implementation dp[n][k] represents
    // maximum sum subsequence of length k and the
    // subsequence is ending at index n.
    int dp[n][k + 1], ans = -1;
  
    // Initializing whole multidimensional 
    // dp array with value -1
    memset(dp, -1, sizeof(dp));
  
    // For each ith position increasing subsequence
    // of length 1 is equal to that array ith value
    // so initializing dp[i][1] with that array value
    for (int i = 0; i < n; i++) {
        dp[i][1] = arr[i];
    }
  
    // Starting from 1st index as we have calculated
    // for 0th index. Computing optimized dp values 
    // in bottom-up manner
    for (int i = 1; i < n; i++) {
        for (int j = 0; j < i; j++) {
  
            // check for increasing subsequence
            if (arr[j] < arr[i]) {
                for (int l = 1; l <= k - 1; l++) {
  
                    // Proceed if value is pre calculated
                    if (dp[j][l] != -1) {
  
                        // Check for all the subsequences 
                        // ending at any j<i and try including
                        // element at index i in them for
                        // some length l. Update the maximum 
                        // value for every length.
                        dp[i][l + 1] = max(dp[i][l + 1], 
                                          dp[j][l] + arr[i]);
                    }
                }
            }
        }
    }
  
    // The final result would be the maximum
    // value of dp[i][k] for all different i.
    for (int i = 0; i < n; i++) {
        if (ans < dp[i][k])
            ans = dp[i][k];
    }
  
    // When no subsequence of length k is 
    // possible sum would be considered zero
    return (ans == -1) ? 0 : ans;
}
  
// Driver function
int main()
{
    int n = 8, k = 3;
    int arr[n] = { 8, 5, 9, 10, 5, 6, 21, 8 };
    int ans = MaxIncreasingSub(arr, n, k);
    cout << ans << "\n";
    return 0;
}
chevron_right

filter_none

edit
close

play_arrow

link
brightness_4
code

/*Java program to calculate the maximum sum of 
increasing subsequence of length k*/
import java.util.*;
  
class GFG
{
      
static int MaxIncreasingSub(int arr[], int n, int k) 
    // In the implementation dp[n][k] represents 
    // maximum sum subsequence of length k and the 
    // subsequence is ending at index n. 
    int dp[][]=new int[n][k + 1], ans = -1
  
    // Initializing whole multidimensional 
    // dp array with value -1 
    for(int i = 0; i < n; i++)
        for(int j = 0; j < k + 1; j++)
            dp[i][j]=-1;
  
    // For each ith position increasing subsequence 
    // of length 1 is equal to that array ith value 
    // so initializing dp[i][1] with that array value 
    for (int i = 0; i < n; i++) 
    
        dp[i][1] = arr[i]; 
    
  
    // Starting from 1st index as we have calculated 
    // for 0th index. Computing optimized dp values 
    // in bottom-up manner 
    for (int i = 1; i < n; i++) 
    
        for (int j = 0; j < i; j++) 
        
  
            // check for increasing subsequence 
            if (arr[j] < arr[i]) 
            
                for (int l = 1; l <= k - 1; l++) 
                
  
                    // Proceed if value is pre calculated 
                    if (dp[j][l] != -1)
                    
  
                        // Check for all the subsequences 
                        // ending at any j<i and try including 
                        // element at index i in them for 
                        // some length l. Update the maximum 
                        // value for every length. 
                        dp[i][l + 1] = Math.max(dp[i][l + 1], 
                                        dp[j][l] + arr[i]); 
                    
                
            
        
    
  
    // The final result would be the maximum 
    // value of dp[i][k] for all different i. 
    for (int i = 0; i < n; i++) 
    
        if (ans < dp[i][k]) 
            ans = dp[i][k]; 
    
  
    // When no subsequence of length k is 
    // possible sum would be considered zero 
    return (ans == -1) ? 0 : ans; 
  
// Driver code 
public static void main(String args[])
    int n = 8, k = 3
    int arr[] = { 8, 5, 9, 10, 5, 6, 21, 8 }; 
    int ans = MaxIncreasingSub(arr, n, k); 
    System.out.println(ans ); 
      
  
// This code is contributed by Arnab Kundu
chevron_right

filter_none

edit
close

play_arrow

link
brightness_4
code

# Python program to calculate the maximum sum
# of increasing subsequence of length k
  
def MaxIncreasingSub(arr, n, k):
      
    # In the implementation dp[n][k] represents 
    # maximum sum subsequence of length k and the 
    # subsequence is ending at index n.
    dp = [-1]*n
    ans = -1
  
    # Initializing whole multidimensional
    # dp array with value - 1
    for i in range(n):
        dp[i] = [-1]*(k+1)
  
    # For each ith position increasing subsequence
    # of length 1 is equal to that array ith value
    # so initializing dp[i][1] with that array value
    for i in range(n):
        dp[i][1] = arr[i]
      
    # Starting from 1st index as we have calculated
    # for 0th index. Computing optimized dp values
    # in bottom-up manner
    for i in range(1,n):
        for j in range(i):
              
            # check for increasing subsequence
            if arr[j] < arr[i]:
                for l in range(1,k):
  
                    # Proceed if value is pre calculated
                    if dp[j][l] != -1:
                          
                        # Check for all the subsequences
                        # ending at any j < i and try including
                        # element at index i in them for
                        # some length l. Update the maximum
                        # value for every length.
                        dp[i][l+1] = max(dp[i][l+1],
                                        dp[j][l] + arr[i])
      
    # The final result would be the maximum 
    # value of dp[i][k] for all different i.
    for i in range(n):
        if ans < dp[i][k]:
            ans = dp[i][k]
      
    # When no subsequence of length k is
    # possible sum would be considered zero
    return (0 if ans == -1 else ans)
  
# Driver Code
if __name__ == "__main__":
  
    n, k = 8, 3
    arr = [8, 5, 9, 10, 5, 6, 21, 8]
    ans = MaxIncreasingSub(arr, n, k)
    print(ans)
  
# This code is contributed by
# sanjeev2552
chevron_right

filter_none

edit
close

play_arrow

link
brightness_4
code

/*C# program to calculate the maximum sum of 
increasing subsequence of length k*/
using System;
  
class GFG
{
      
static int MaxIncreasingSub(int []arr, int n, int k) 
    // In the implementation dp[n,k] represents 
    // maximum sum subsequence of length k and the 
    // subsequence is ending at index n. 
    int [,]dp=new int[n, k + 1];
    int ans = -1; 
  
    // Initializing whole multidimensional 
    // dp array with value -1 
    for(int i = 0; i < n; i++)
        for(int j = 0; j < k + 1; j++)
            dp[i, j]=-1;
  
    // For each ith position increasing subsequence 
    // of length 1 is equal to that array ith value 
    // so initializing dp[i,1] with that array value 
    for (int i = 0; i < n; i++) 
    
        dp[i, 1] = arr[i]; 
    
  
    // Starting from 1st index as we have calculated 
    // for 0th index. Computing optimized dp values 
    // in bottom-up manner 
    for (int i = 1; i < n; i++) 
    
        for (int j = 0; j < i; j++) 
        
  
            // check for increasing subsequence 
            if (arr[j] < arr[i]) 
            
                for (int l = 1; l <= k - 1; l++) 
                
  
                    // Proceed if value is pre calculated 
                    if (dp[j, l] != -1)
                    
  
                        // Check for all the subsequences 
                        // ending at any j<i and try including 
                        // element at index i in them for 
                        // some length l. Update the maximum 
                        // value for every length. 
                        dp[i, l + 1] = Math.Max(dp[i, l + 1], 
                                        dp[j, l] + arr[i]); 
                    
                
            
        
    
  
    // The final result would be the maximum 
    // value of dp[i,k] for all different i. 
    for (int i = 0; i < n; i++) 
    
        if (ans < dp[i, k]) 
            ans = dp[i, k]; 
    
  
    // When no subsequence of length k is 
    // possible sum would be considered zero 
    return (ans == -1) ? 0 : ans; 
  
// Driver code 
public static void Main(String []args)
    int n = 8, k = 3; 
    int []arr = { 8, 5, 9, 10, 5, 6, 21, 8 }; 
    int ans = MaxIncreasingSub(arr, n, k); 
    Console.WriteLine(ans ); 
}
  
// This code contributed by Rajput-Ji
chevron_right

Output:
40

Time complexity: O(n^2*k)
Space complexity: O(n^2)




Competitive Programmer, Full Stack Developer, Technical Content Writer, Machine Learner

If you like GeeksforGeeks and would like to contribute, you can also write an article using contribute.geeksforgeeks.org or mail your article to contribute@geeksforgeeks.org. See your article appearing on the GeeksforGeeks main page and help other Geeks.

Please Improve this article if you find anything incorrect by clicking on the "Improve Article" button below.



Improved By : andrew1234, Rajput-Ji, sanjeev2552



Article Tags :