Interview Note - Dynanmic Programming

Jan 25 2015 (Jan 25 2015) English>Interview 8 minutes read (About 1184 words)

1. Dynamic Programming

A method for solving a complex problem by breaking it down into a collection of simpler sub-problems.

1.1 When to use

One of the following three:

Maximum/Minimum Problem
Yes or No Question
Count all possible solutions

Can’t do sort or swap operation

Note: DP can’t return all results, it only returns max/min, yes/no or a certain value(like length, possible solutions, etc.).

1.2 How to think

States: what we need to store for each sub-problem (usually an array).
Function: what is the relationship between each state.
Intialization: what is the start of each state.
Answer: what is the end of each state.

Example: Number Triangle
We use an array f[i][j] to record the minimum sum from (0, 0) to (i, j). For a certain (i, j), we know it is either from (i - 1, j - 1) or (i - 1, j). So the sub-problem is to traverse all elements in the triangle and calculate all the f[i][j] = min(f[i - 1][j - 1], f[i - 1][j]) + T[i]][j]. The result will be the minimum one in f[n - 1][0, ..., n - 1].
We can also do it from bottom up. First we find the minimum of each pair in the last row, then we do above similarly but from bottom up. Finally the only one element left in the array is the result we are looking for.
A bonus point is doing this using only O(n) extra space, where n is the total number of rows in the triangle. Since each time we calculate f[i][j], the result is only related with f[i - 1][], we can only use 1D array and update the elements in it in each for loop.
Here is the code from bottom up with O(n) space cost.

public class Solution {
    public int minimumTotal(List<List<Integer>> triangle) {
        if(triangle.size() == 0) {
            return 0;
        }

        int[] result = new int[triangle.size() + 1];
        
        for(int row = triangle.size() - 1; row >= 0; row--) {
            for(int col = 0; col <= row; col++) {
                result[col] = Math.min(result[col], result[col + 1]) + triangle.get(row).get(col);
            }
        }
        
        return result[0];
    }
}

Dynamic Programming uses extra space to remember the mid-result. So that it is more efficient than recursive searching, which may repeated calculate the same mid-result for many times. In the above example, if we do a brute force search, it cost O(2^n) time, where n is the height of the triangle (each row, we need to choose one from two, totally n rows). But it only costs O(n^2) in theoretical(m rows, each row has n elements, totally n^2) or O(n), which means all numbers we only need to visit once.

Interview Note - Linked List

Jan 24 2015 (Jan 24 2015) English>Interview 6 minutes read (About 857 words)

1. Introduce Dummy Node

When the head of the target list we want to return may be different from the original given list, we can use a dummy node linked to the result so that we can get what we want by calling dummy.next.

Remove Duplicates from Sorted List II:

public class Solution {
    public static ListNode deleteDuplicates(ListNode head) {
        if(head == null || head.next == null) {
            return head;
        }

        ListNode dummy = new ListNode(0);
        ListNode pre = dummy;
        ListNode cur = head;
        int dup = Integer.MIN_VALUE;

        while(cur != null && cur.next != null) {
            if(cur.val == cur.next.val) {
                dup = cur.val;
                while(cur.val == dup) {
                    cur = cur.next;
                    // if the rest of elements are all the same
                    if(cur == null) {
                        return dummy.next;
                    }
                }
            } else {
                pre.next = cur;
                pre = pre.next;
                cur = cur.next;
                // add null to pre.next: in case the rest of elements are the same and returned above
                pre.next = null;
            }
        }
        // add the last element (since cur.next != null skip this element)
        pre.next = cur;

        return dummy.next;
    }
}

Reverse Linked List II:

public class Solution {
    public ListNode reverseBetween(ListNode head, int m, int n) {
        if(head == null) {
            return head;
        }

        // find the m position node
        int count = 1;
        ListNode dummy = new ListNode(0);
        dummy.next = head;
        ListNode pre = dummy;
        while(count < m) {
            pre = pre.next;
            count++;
        }

        // do the reverse: pre-> reverse[first->cur->next] ...
        ListNode first = pre.next;
        ListNode cur = pre.next.next;
        while(count < n) {
            ListNode next = cur.next;
            cur.next = pre.next;
            pre.next = cur;
            cur = next;
            count++;
            if(count == n) {
                first.next = next;
            }
        }

        return dummy.next;
    }
}

Interview Note - Binary Tree

Jan 23 2015 (Jan 23 2015) English>Interview 6 minutes read (About 953 words)

1. Binary Tree DFS Traversal

1.1 Recursion

This method will cost O(n) time with no extra space. The space is assigned by system(which could be O(1) or O(h), h is the height of the tree).

public ArrayList<Integer> preorderTraversal(TreeNode root) {
    ArrayList<Integer> result = new ArrayList<Integer>();
    helper(root, result);
    return result;
}

private void helper(TreeNode root, ArrayList<Integer> result) {
    if(root == null) {
        return;
    }
    result.add(root.val);
    helper(root.left, result);
    helper(root.right, result);
}

1.2 Iteration

Use a stack to store the nodes. Pay attention to the order of storing. We store the right subtree node before the left one, so that we can get the preorder nodes when pop from the stack.
This method will cost O(n) time with O(h) extra space for the stack.

 */
public class Solution {
    public ArrayList<Integer> preorderTraversal(TreeNode root) {
        ArrayList<Integer> result = new ArrayList<Integer>();
        if(root == null) {
            return result;
        }
        
        Stack<TreeNode> stack = new Stack<TreeNode>();
        stack.push(root);
        while(!stack.isEmpty()) {
            TreeNode cur = stack.pop();
            result.add(cur.val);
            if(cur.right != null) {
                stack.push(cur.right);
            }
            if(cur.left != null) {
                stack.push(cur.left);
            }
        }
        
        return result;
    }
}

Interview Note - Binary Search & Sorted Array

Jan 22 2015 (Jan 22 2015) English>Interview 6 minutes read (About 854 words)

Binary Search

For a given sorted array (ascending order) and a target number, find the first index of this number in O(log n) time complexity.
If the target number does not exist in the array, return -1.
Example:
If the array is [1, 2, 3, 3, 4, 5, 10], for given target 3, return 2.

class Solution {
    public int binarySearch(int[] nums, int target) {
        if(nums == null || nums.length == 0) {
            return -1;
        }

        int start = 0;
        int end = nums.length - 1;
        while(start < end) {
            int mid = start + (end - start) / 2;
            if(nums[mid] >= target) {
                end = mid;
            } else {
                start = mid + 1;
            }
        }

        return nums[start] == target ? start : -1;
    }
}

Key Points:

while(start < end): when start and end are overlapped each other, break. At this moment, we can return start or end. But it is different when we need to return the index before or after the target(see point 3).
int mid = start + (end - start) / 2: in order to avoid overflow of start + end, we need to do a little trick here.
while loop: in the while loop, there are two different needs:

return the index of the target: as the example above.

return the one before(example below)/after the target:
Search Insert Position

while(start <= end) {
    int mid = start + (end - start) / 2;
    if(A[mid] == target) {
        return mid;
    } else if(A[mid] < target) {
        start = mid + 1;
    } else {
        end = mid - 1;
    }
}

Here, we need to use while(start <= end) so that when we break the loop, two pointers are at two different(but neighbored) positions. Besides, if we want to return the one before the index, we return start, otherwise we return end.

Interview Note - Intro

Jan 21 2015 (Jan 21 2015) English>Interview 3 minutes read (About 393 words)

Coding Style

Leave blank space around each operations
Give meaningful names to parameters
Add space line between two logic blocks

For example, Implement strStr():

public class Solution {
    public String strStr(String haystack, String needle) {
        if(haystack.length() < needle.length()) {
            return null;
        }
        if(haystack == null || needle == null || needle.length() == 0) {
            return haystack;
        }

        for(int i = 0; i <= haystack.length() - needle.length(); i++) {
            boolean find = true;
            for(int j = 0; j < needle.length(); j++) {
                if(haystack.charAt(i + j) != needle.charAt(j)) {
                    find = false;
                    break;
                }
            }
            if(find == true) {
                return haystack.substring(i);
            }
        }

        return null;
    }
}

Tips: A well-known algorithm to solve this problem is using KMP. This method can run in O(m+n) time. But usually, during a 45 minutes interview, complete this O(m*n) method is enough. So, understanding what the point the interviewer asks is very important.

Think it in this way. If I were the interview:
This guy may be my potential colleague. Is there any bug here? Because if there are many bugs, I will probably debug a lot for him when we are in a team.
Is he talktive? Well, I will give you limited conditions about the problem and I hope you can ask me for more details. Also, to every question, pay attention to the boundry condition.

When we solve a coding problem

Pay attention to the boundry condition
Do communication during the writing part
When finished coding, do a test case no matter the interviewer request or not

Permutations Problems

Subsets:

public class Solution {
    public ArrayList<ArrayList<Integer>> subsets(int[] S) {
        ArrayList<ArrayList<Integer>> result = new ArrayList<ArrayList<Integer>>();
        ArrayList<Integer> subset = new ArrayList<Integer>();
        Arrays.sort(S);
        helper(result, subset, 0, S);
        return result;
    }
    
    private void helper(ArrayList<ArrayList<Integer>> result, ArrayList<Integer> subset, int len, int[] S) {
        result.add(new ArrayList<Integer>(subset));
        for(int i = len; i < S.length; i++) {
            subset.add(S[i]);
            helper(result, subset,i + 1, S);
            subset.remove(subset.size() - 1);
        }
    }
}

#InterviewNote

Interview Note - Dynanmic Programming

1. Dynamic Programming

1.1 When to use

1.2 How to think

Interview Note - Linked List

1. Introduce Dummy Node

Interview Note - Binary Tree

1. Binary Tree DFS Traversal

1.1 Recursion

1.2 Iteration

Interview Note - Binary Search & Sorted Array

Binary Search

Interview Note - Intro

Coding Style

When we solve a coding problem

Permutations Problems

Your browser is out-of-date!