Maximum Gap

Given an unsorted array, find the maximum difference between the successive elements in its sorted form. 

 Try to solve it in linear time/space. 

 Return 0 if the array contains less than 2 elements.

 You may assume all elements in the array are non-negative integers and fit in the 32-bit signed integer range.

In order to solve this in linear time, we cannot use general sorting algorithm. Here we use bucket sort.

  • First, we scan the array to find the min and the max values.  
  • We can then create n - 1 buckets with equal size. Suppose we have n elements in total, then we have n - 2 elements except the min and the max. Scan the array again to put the n - 2 elements into these buckets. We should have at least one empty bucket. For each bucket, we record the min and the max for all the elements within that bucket. 
  • Finally, we scan all the buckets, and maintain the maximum gap. Note that the maximum gap between the successive elements in the sorted array can only be the gap between buckets. Why? Try to consider the size of each bucket.
Therefore, the total time complexity is O(3n) and the space complexity is O(2n).

Comments

  1. AnonymousDecember 14, 2014 at 7:22 PM

    maxdiff = max(maxdiff, it.first - lastval); would you explain this? I am understanding, in each bucket, first is min, second is max, so in order to get max gap, you would like to do it.second - lastMin, but here it looks like you are doing it.first - lastMax. Why?

    ReplyDelete
    Replies
    1. UnknownDecember 14, 2014 at 9:34 PM

      Note that the size of each bucket is (max - min) / (n - 1), which is less than or equal to the maximum gap (since we have n - 1 gaps for n numbers, and the total length is max - min). Therefore, the maximum gap can only be the gap between the maximum element in the i-th bucket and the minimum element in the (i+1)-th bucket.

      Delete
    2. UnknownDecember 14, 2014 at 9:46 PM

      An example is, [4, 1, 8, 3, 10, 7]
      min is 1 and max is 10, the gap is 9/5 = 1.8
      so the five buckets are
      [1, 2.8), is empty (since 1 is minimum and has been ignored)
      [2.8, 4.6), min is 3 and max is 4
      [4.6, 6.4), is empty
      [6.4, 8.2), min is 7 and max is 8
      [8.2, 10), is empty

      Then we scan these buckets again. If we ignore the empty buckets, it makes sense to compare the following gaps:
      3 (min of the second bucket) - 1 (minimum in total)
      7 (min of the fourth bucket) - 4 (max of the second bucket)
      10 (maximum in total) - 8 (max of the fourth bucket)

      and we found that the maximum gap is 3.

      Delete

Post a Comment

Popular posts from this blog

[ITint5] Maximum Subarray for a Circular Array

http://www.itint5.com/oj/#8 Classic DP problem, which has also appeared in Leetcode as Maximum subarray . The only difference is, in this problem, empty subarray (where sum is 0) is allowed. So instead of return ret, we could simply return max(0, ret). http://www.itint5.com/oj/#9 Calculate the maximum subarray for a circular array. For example, [1, 3, -2, 6, -1], we should return 9 since 6 + (-1) + 1 + 3 = 9. The following solution might not be optimized. Suppose that the maximum subarray is from A[i] to A[j]. The idea is, consider the following different scenarios: i <= j: it is the same as classic problem. i > j: the sum is A[i : end] + A[start : j]. So we can use two additional arrays: left[j] = maximum subarray, starts from A[start] and ends no later than A[j] right[i] = maximum subarray, starts no earlier than A[i] and ends at A[end] The following code passes the online judge. However, we need O(4n) in time and O(2n) in space. Not sure how to improve this.
Read more

[CC150] Chapter 4 Trees and Graphs

4.1 Implement a function to check if a binary tree is balanced. For the purpose of this question, a balanced tree is defined to be a tree such that the heights of the two subtrees of any node never differ by more than one. This problem is in leetcode as  Balanced Binary Tree . 4.2 Given a directed graph, design an algorithm to find out whether there is a route between two nodes. Both BFS and DFS works. Here I use iterative BFS. 4.3 Given a sorted (increasing order) array, write an algorithm to create a binary search tree with minimal height. This problem is in leetcode as  Convert Sorted Array To Binary Search . Related problem:  Convert Sorted List To Binary Search . 4.4 Given a binary search tree, design an algorithm which creates a linked list of all the nodes at each depth (e.g., if you have a tree with depth D, you will have D linked lists). This problem is equivalent to  Binary Tree Level Order Traversal . Related problem:  Binary Tree Zigzag Level Order Traversal
Read more