Biggest valued, 'H' shaped area in matrix? - c++

Question

I got the task where I must find the H shaped region which has the biggest sum of numbers in it. Under 'H' shaped region, tha task meant this, consisting of 7 elements and never changing:
x x
xxx
x x
The matrix's size must be 33 or bigger than that, and I don't have to work with rotated 'H' shape. However, it can move upwards and downwards if the matrix is that big (for example a 46 matrix).
I thought of first counting a "maximum" value, starting from the [0][0] element. However, I can't figure out how to move this region-counting along. Could you help me out, please?
Here's my code so far:
#include<iostream>
int main(){
int n = 3;
int m = 4;
int mtx[n][m] = {
1,1,1,3,
1,1,1,3,
1,1,1,3
};
//counting the maximum H value
int max = 0;
for(int i = 0; i < n; i++){
max += mtx[i][0];
}
for(int i = 0; i < n; i++){
max += mtx[i][2];
}
max += mtx[1][1];
int counter = 0;
int j = 0;
int k = 0;
//finding if there is bigger
while(counter >max){
//questioned area, not sure what to do here
if(counter < max){
max = counter;
}
}
return 0;
}

Answer 1

As mentioned in a comment, knowing the single maximum element in the matrix does not help to find the maximum H shape. A concrete counter example:
0 1 0 5
9 1 1 5
0 1 0 5
Maximum element is 9 but maximum sum H shape is
1 0 5
1 1 5
1 0 5
You would need to add more information to know whether the maximum element is part of the maximum H: Only if max_element + 6*min_element > 7*second_smallest_element you can be sure that max_element is part of the biggest sum H. This condition could be refined, but it cannot be made such that the biggest element is always part of the biggest sum H, because thats not true in general (see counter example above).
As suggested in another comment, you should write a function that given coordinates of the upper left corner calculates the sum of elements in the H shape:
#include <iostream>
#include <array>
int H_sum(const std::array<std::array<int,4>,3>& matrix, int x0,int y0){
// A E
// BDF
// C G
int sum = matrix[x0][y0]; // A
sum += matrix[x0+1][y0]; // B
sum += matrix[x0+2][y0]; // C
sum += matrix[x0+1][y0+1]; // D
sum += matrix[x0][y0+2]; // E
sum += matrix[x0+1][y0+2]; // F
sum += matrix[x0+2][y0+2]; // G
return sum;
}
int main() {
std::array<std::array<int,4>,3> mtx{
1,1,1,3,
1,1,1,3,
1,1,1,3
};
for (const auto& row : mtx){
for (const auto& e : row){
std::cout << e;
}
std::cout << "\n";
}
std::cout << H_sum(mtx,0,0);
}
This is of course only something to get you started. Next you have to carefully consider what are the maximum indices you can pass to H_sum without going out-of-bounds. Then write a nested loop to scan all positions of the H and remember the maximum value encountered.
Last but not least, what I described so far is a brute force approach. You calculate sum for all possible H shapes, remember the maximum, and are done. Maybe there is a clever trick to avoid adding all elements multiple times (for example in a larger matrix, the right leg of one H is the left leg of a different H). Though before applying such tricks and trying to be clever I strongly suggest to write something that is perhaps slow but correct, easy to read and verify.