[Java]{DP}(With Explanation)

  • 1

    Very Similar to Longest Common Subsequence Problem.

    Let, s1 & s2 be the two strings with 1 based indexes.
    Now assume, dp[i][j] = minimumDeleteSum( s1[0,i], s2[0,j])

    Base case:
    When either of the strings is empty, then whole of the other string has to be deleted.
    for e.g. if s1 = "", s2 = "abc", then only way we could match these strings by deleting characters is by dropping 'a','b','c' of s2 to make it empty like s1.

    Thus, whenever one of them is empty(i.e. i==0 or j==0) then answer is sum of ASCII code of the characters of the other string.

    Hence the 1st rule: dp[i][j] =

    • sum_ascii(s2) -> if i==0
    • sum_ascii(s1) -> if j==0

    Non-Base case

    Of the two strings, if both of their last characters match then certainly the answer comes from skipping those characters.
    i.e. Answer("zca","bza") = Answer("zc","bz")

    Hence the 2nd rule: dp[i][j] =

    • dp[i-1][j-1] -> if s1[i]==s2[j]

    Finally, if the last characters are different then its one of the three situations:

    • drop s1's last character (ASCII(s1's last) + dp[i-1][j])
    • drop s2's last character (ASCII(s2's last) + dp[i][j-1])
    • drop both last characters (ASCII(s1's last) + ASCII(s2's last) + dp[i-1[[j-1])

    Hence the 3rd rule: dp[i][j] =

    • Min((ASCII(s1's last) + dp[i-1][j]),(ASCII(s2's last) + dp[i][j-1]),(ASCII(s1's last) + ASCII(s2's last) + dp[i-1[[j-1]))

    Combining these 3 rules gives us an elegant solution.

    public int minimumDeleteSum(String s1, String s2) {
            int m = s1.length();
            int n = s2.length();
            int[][] dp = new int[m+1][n+1];
            for(int i=0;i<=m;i++){
                for(int j=0;j<=n;j++){
                    if(i==0 || j==0){
                        int a = 0;
                        for(int z=1;z<=Math.max(j,i);z++){
                            a += (i==0?s2.charAt(z-1):s1.charAt(z-1));
                        dp[i][j] = a;
                    else if(s1.charAt(i-1)==s2.charAt(j-1)){
                        dp[i][j] = dp[i-1][j-1];
                        dp[i][j] = Math.min(s1.charAt(i-1)+dp[i-1][j],s2.charAt(j-1)+dp[i][j-1]);
                        dp[i][j] = Math.min(dp[i][j],s1.charAt(i-1)+s2.charAt(j-1)+dp[i-1][j-1]);
            return dp[m][n];

  • 0

    @thalaiva Excellent. But shoudnt 3rd Rule be Min(...) instead of Max(...)

  • 0

    @khader Yes, you are right! My bad. Thank you for pointing out the mistake!

Log in to reply

Looks like your connection to LeetCode Discuss was lost, please wait while we try to reconnect.