Finding Max value in an array using recursion

It's actually much simpler than that. The base case is if you've reached the end of the array (the 'else' part of the ternary control block below). Otherwise you return the max of the current and the recursive call.

public static int findMax(int[] a) {
    return findMax(a, 0);
}
private static int findMax(int[] a, int i) {
    return i < a.length
           ? Math.max(a[i], findMax(a, i + 1))
           : Integer.MIN_VALUE;
}

At each element, you return the larger of the current element, and all of the elements with a greater index. Integer.MIN_VALUE will be returned only on empty arrays. This runs in linear time.

I came across this thread and it helped me a lot. Attached is my complete code in both recursion and divide&conquer cases. The run time for divide&conquer is slightly better than recursion.

//use divide and conquer.
public int findMaxDivideConquer(int[] arr){
    return findMaxDivideConquerHelper(arr, 0, arr.length-1);
}
private int findMaxDivideConquerHelper(int[] arr, int start, int end){
    //base case
    if(end - start  <=  1) return Math.max(arr[start], arr[end]);
    //divide
    int mid = start + ( end - start )/2;
    int leftMax =findMaxDivideConquerHelper(arr, start, mid);
    int rightMax =findMaxDivideConquerHelper(arr, mid+1, end);
    //conquer
    return Math.max( leftMax, rightMax );
}

// use recursion. return the max of the current and recursive call
public int findMaxRec(int[] arr){
    return findMaxRec(arr, 0);
}
private int findMaxRec(int[] arr, int i){
    if (i == arr.length) {
        return Integer.MIN_VALUE;
    }
    return Math.max(arr[i], findMaxRec(arr, i+1));
}

I would solve this by dividing the array in to the half on each recursive call.

 findMax(int[] data, int a, int b)

where a and b are array indices.

The stop condition is when b - a <= 1, then they are neighbours and the max is max(a,b);

The initial call:

 findMax(int[] data, int 0, data.length -1);

This reduces the maximum recursion depth from N to log2(N).
But the search effort still stays O(N).

This would result in

int findMax(int[] data, int a, int b) {
   if (b - a <= 1) {
     return Math.max(data[a], data[b]);
   } else {
     int mid = (a+b) /2; // this can overflow for values near Integer.Max: can be solved by a + (b-a) / 2; 
     int leftMax =  findMax(a, mid);
     int rightMax = findMax(mid +1, b);
     return Math.max(leftMax, rightMax);
   }
}

You could just as easily do it with only one counter, just the index of the value you want to compare this time:

public static int findMax(int[] a, int index) {
    if (index > 0) {
        return Math.max(a[index], findMax(a, index-1))
    } else {
        return a[0];
    }
}

This much better shows what is going on, and uses the default "recursion" layout, e.g. with a common base step. Initial call is by doing findMax(a, a.length-1).