Ziyuan Qin's blog

Super Egg Drop

RecursionDP
https://leetcode.com/problems/super-egg-drop

# Solution

See this video in Chinese (opens new window) for perfect explanation on this problem and the previous easier problem Drop Eggs.

Let ii be the one of K eggs, and jj be one of the N floors.

# Iterative DP (TLE)

The state transition is:

dp(i,j)=min⁡1≤k≤j(max⁡(dp(i−1,k−1),dp(i,j−k)))dp(i, j) = \min_{1 \le k \le j}\left(\max(dp(i-1, k-1), dp(i, j-k)) \right)

If ith egg breaks at kth floor, there are i-1 eggs left and critical floor exists below k, so problem is reduced to res[i-1][k-1].

Else, ith egg doesn't break and critical floor exists btw k and j, so problem is reduced to res[i][j-k].

Note:

  • plus 1 for initial drop
  • max takes into account the worst case
  • min takes the number of drops for the best way to drop ith egg at optimal kth floor

Complexity

time: O(KN2)O(KN^2)
space: O(KN)O(KN)

def superEggDrop(self, K, N):
    res = [[sys.maxsize for _ in range(N+1)] for _ in range(K+1)]

    for i in range(1, K+1):
        res[i][0] = 0
        res[i][1] = 1
    for j in range(1, N+1):
        res[1][j] = j
    for i in range(2, K+1):
        for j in range(2, N+1):
            for k in range(1, j+1):
                res[i][j] = min(res[i][j], 1 + max(res[i-1][k-1], res[i][j-k]))
    return res[m][n]
1
2
3
4
5
6
7
8
9
10
11
12
13

# Recursive DP w/ Memoization (TLE)

def superEggDrop(self, K, N):
    if N == 0 or N == 1:
        return N
    if K == 1:
        return N
    min = sys.maxsize

    # Consider all droppings from 1st floor to kth floor
    # and return the minimum of these values plus 1
    for x in range(1, N + 1):
        res = 1 + max(self.superEggDrop(K - 1, x - 1), self.superEggDrop(K, N - x))
        if (res < min):
            min = res
    return min
1
2
3
4
5
6
7
8
9
10
11
12
13
14

# DP w/ Optimality Criterion (REDO)

def superEggDrop(self, K, N):
    # Right now, dp[i] represents dp(1, i)
    dp = range(N+1)

    for k in range(2, K+1):
        # Now, we will develop dp2[i] = dp(k, i)
        dp2 = [0]
        x = 1
        for n in range(1, N+1):
            # Let's find dp2[n] = dp(k, n)
            # Increase our optimal x while we can make our answer better.
            # Notice max(dp[x-1], dp2[n-x]) > max(dp[x], dp2[n-x-1])
            # is simply max(T1(x-1), T2(x-1)) > max(T1(x), T2(x)).
            while x < n and max(dp[x-1], dp2[n-x]) > max(dp[x], dp2[n-x-1]:
                x += 1

            # The final answer happens at this x.
            dp2.append(1 + max(dp[x-1], dp2[n-x]))

        dp = dp2

    return dp[-1]
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

# Math (REDO)

def superEggDrop(self, K, N):
    dp = [[0 for col in range(K + 1)] for row in range(N + 1)]
    m = 0
    while dp[m][K] < N
        m = m + 1
        for i in range(1,K + 1)
            dp[m][i] = dp[m - 1][i - 1] + dp[m - 1][i] + 1
    return m
1
2
3
4
5
6
7
8

Symmetric Tree