Unique Binary Search Trees | LeetCode | DP Solution, Explained

LeetCode 96. Unique Binary Search Trees. I explain the Dynamic Programming Solution in detail, along with intuitions on how to actually think about this problem. Problem Link.

0. Unique Binary Search Trees: Problem Discussion

0.0. I/O format

• Input: n
• Return: The number of unique BSTs

0.1. What are BSTs?

• The existence of nodes x and z is not necessary, but if they do (any of them), they should follow the condition x < y < z

1. Unique Binary Search Trees: Observations & Logic

1.0. Intuitions

In these type of counting problems, I like to aim for finding a pattern between the first few cases – maybe the i+1th answer is double of ith answer, or maybe there is a common difference between adjacent answers.

Its a good idea to look at the couple of smaller cases first, since they are easy to do manually, and in dp questions, they become a bliss. Let’s see what happens!

1.1. Observations

Huh. Maybe for n=x –> x? Let’s see a couple more examples to confirm or deny this.

Ah. For n = x –> x will not work. There’s something more going on. Look at the root of the trees for n = 3.

Interestingly enough, it looks as if there are some repeating sub-structures in the case of n = 3. Look at node 3 as an example. The nodes 1 and 2 will both be on the left side, and both have 2 arrangements they can be in – since n = 2 –> 2!

This means we are looking at a DP question.

In this case, we group the elements that would have come to the right and left side (using the property of BST), and we will call the appropriate n = 2 –> 2 OR n = 3 –> 5.

2. Unique Binary Search Trees: Implementation

The implementation is straightforward once we understand the DP logic:

• calculate some base cases, dp[0] = dp[1] = 1
• where dp[i] represents the answer for the ith node, as in, n = i –> dp[i]
• then, for each i, we try to find all the possible arrangements, each time increasing the left_side count by 1, and decreasing the right_side count by 1 simultaneously.

2.0. Code

class Solution:
    def numTrees(self, n: int) -> int:
        dp = [0]*(n+1)
        dp[0] = 1
        dp[1] = 1
        
        for i in range(2, n+1):
            dp[i] = 0
            
            l, r = 0, i-1
            while l < i:
                dp[i] += dp[l]*dp[r]
                l += 1
                r -= 1
            
        return dp[n]

2.1. Complexity Analysis

• Time: O(N²). O(N) to iterate over all the Ns and another O(N) nested to consider all left and right combinations.
• Space: O(N). To store all the dp values.

Tanishq Chaudhary

Producing high-quality intuitive explanations of interview problems. Currently covering LeetCode and InterviewBit.