Maximum subsequence sum of at most K-distant adjacent elements

Given an array arr[] consisting of integers of length N and an integer K (1 <= K <= N), the task is to find the maximum sub-sequence sum in the array such that adjacent elements in that sub-sequence have at most a difference of K in their indices in the original array. In other words, if i and j are indices of two consecutive elements of sub-sequence in original array then |i-j| <= K  .


Input: arr[] = {1, 2, -2, 4, 3, 1}, K = 2
Output: 11
Explanation : The sub
sequence with maximum sum is {1, 2, 4, 3, 1} (difference between indices <=2)

Input: arr[] = {4, -2, -2, -1, 3, -1}, K = 2
Output: 5
Explanation : The sub-sequence with maximum sum is {4, -2, 3} (difference between indices <=2)

Naive approach: Generate all possible subsets of the array and check for each of the subsets whether it satisfies the condition such that two adjacent elements have at most difference of K in their indices. If yes, then compare its sum with the largest sum obtained till now and update the sum if it is greater than the obtained sum till now.

Efficient Approach: This problem can be solved using Dynamic Programming

Create a table dp[], where dp[i] stores the largest possible sum for the sub-sequence till the ith index.

  • If the current element is the first element of the sub-sequence then:

   dp[i] =arr[i]

  • Otherwise, we have to check previous results and check what is the maximum result of dp in a window of K behind this index :

   dp[i] = max(arr[i] + dp[x]) where x is from [i-k, i-1]


  • For every index choose that condition which gives the maximum sum at that index, So final recurrence relation will be:

dp[i] = max(arr[i], arr[i] + dp[x])  where   i-k <= x <= i-1

So, for checking what is the maximum value behind this index in a window of K, either we can iterate from dp[i-1] to dp[i-k] and find the maximum value, in which case the time complexity will be O(N*K) or it can be reduced by taking an ordered map and keep maintaining the computed dp[i] values for every index. This reduces the complexity to O(N*log(K)).

Below is the implementation of the above approach.






// C++ implementation to
// C++ program to
// find the maximum sum
// subsequence such that
// two adjacent element
// have atmost difference
// of K in their indices
#include <iostream>
#include <iterator>
#include <map>
using namespace std;
int max_sum(int arr[], int N,
            int K)
    // DP Array to store the
    // maximum sum obtained
    // till now
    int dp[N];
    // Ordered map to store DP
    // values in a window ok K
    map<int, int> mp;
    // Initializing dp[0]=arr[0]
    dp[0] = arr[0];
    // Inserting value of DP[0]
    // in map
    // Intializing final answer
    // with dp[0]
    int ans = dp[0];
    for (int i = 1;
         i < N; i++) {
        // when i<k there is no
        // need to delete elements
        // from map as window is
        // less than K
        if (i < K) {
            // Initializing interator
            // to end af map
            auto it = mp.end();
            // Decreasing iterator to
            // get maximum key
            // Evaluating DP[i]
            // from reccurence
            dp[i] = max(it->first
                            + arr[i],
            // Inserting dp value in map
        else {
            auto it = mp.end();
            dp[i] = max(it->first
                            + arr[i],
            // Deleting dp[i-k] from
            // map beacuse window size
            // has become K+1
            mp[dp[i - K]]--;
            // Erase the key from if
            // count of dp[i-K] becomes
            // zero
            if (mp[dp[i - K]] == 0) {
                mp.erase(dp[i - K]);
        // Calculating final answer
        ans = max(ans, dp[i]);
    return ans;
// Driver code
int main()
    int arr[] = { 1, 2, -2,
                  4, 3, 1 };
    int N = sizeof(arr) / sizeof(int);
    int K = 2;
    cout << max_sum(arr, N, K);
    return 0;




Time Complexity: O(N*log(K))


My Personal Notes arrow_drop_up

Recommended Posts:

Check out this Author's contributed articles.

If you like GeeksforGeeks and would like to contribute, you can also write an article using or mail your article to 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.