My Editorial for Google Kick Start 2021 Round B

Revision en5, by lucifer1004, 2021-04-20 11:12:09

Google Kick Start 2021 Round B Tutorial

中文版 | Chinese version

Here are my solutions to Google Kick Start 2021 Round B. Some of them (C & D) are not optimal, albeit they passed all the test cases.

Problem A — Increasing Substring

We simply go ahead and try appending the current letter after the last letter.

Complexity:

  • Time complexity is $$$\mathcal{O}(|S|)$$$.
  • Space complexity is $$$\mathcal{O}(|S|)$$$.
Code (C++)

Problem B — Longest Progression

For $$$N\leq3$$$, it is obvious that the answer is $$$N$$$.

For $$$N\geq4$$$, we can first calculate $$$l[i]$$$, which denotes the length of the longest arithmetic array from left till $$$i$$$ without making any changes, and $$$r[i]$$$, which denotes the length of the longest arithmetic array from right till $$$i$$$ without making changes.

Then we check each position $$$i$$$ to see if we could get a longer length by changing $$$a[i]$$$ so that we can:

  • combine $$$l[i-1]$$$ and $$$r[i+1]$$$
  • combine $$$l[i-1]$$$ and $$$a[i+1]$$$
  • combine $$$a[i-1]$$$ and $$$r[i+1]$$$

And the complexity:

  • Time complexity is $$$\mathcal{O}(N)$$$.
  • Space complexity is $$$\mathcal{O}(N)$$$.
Code (C++)

Problem C — Consecutive Primes

The official solution leverages the concept of the prime gap and is a very beautiful brute-force solution.

I would not repeat the prime gap part, but will instead talk about the primality test. In this problem, a naive $$$\sqrt{N}$$$ primality test is enough to pass, but we could do better with Miller-Rabin, which runs in $$$\mathcal{O}(K\log^3N)$$$ time.

Since $$$S\leq10^{18}$$$, the largest number we would need to check will be around $$$10^9$$$, in this case, $$$[2,7,61]$$$ would be enough to ensure the correctness of the primality test.

Code (C++)
Original solution (hidden)

Problem D — Truck Delivery

The official solution uses segment tree. However, we can also use blocking to solve this problem.

We separate edges into blocks according to their limits $$$L_i$$$, and each block will have an optimal size of $$$\sqrt{N}$$$.

We first do a depth-first search to gather the information we need. That is, for each node, we would like to know the GCD value of all the edges whose limits are within the same block ($$$acc[i][j]$$$ in the code below), along the path from the root (capital city) to the current node. Also, we would like to know the previous edge ($$$last[i][j]$$$ in the code below) that has a limit that falls within a certain block, along the path.

For each query, we first determine the block $$$j$$$ that $$$W_i$$$ belongs to. Then we can safely calculate the GCD value of all the blocks that are smaller than $$$j$$$. For the $$$j$$$-th block, however, we need to check each edge to find out whether $$$L_k\leq W_i$$$, which can be done with the help of the $$$last$$$ array.

And the complexity:

  • Time complexity is $$$\mathcal{O}((N+Q)\sqrt{N})$$$, if the block size is set as $$$\sqrt{N}$$$. (In the code below, I used a constant block size of $$$500$$$ to avoid discretization.) Also note that all $$$L_i$$$ are distinct, so each block with size $$$\sqrt{N}$$$ can contain at most $$$\sqrt{N}$$$ edges, which ensures the correctness of the time complexity.
  • Space complexity is $$$\mathcal{O}(N\sqrt{N})$$$.
Code (C++)
Tags google, kick start, editorial

History

 
 
 
 
Revisions
 
 
  Rev. Lang. By When Δ Comment
en7 English lucifer1004 2021-04-20 11:43:45 27
en6 English lucifer1004 2021-04-20 11:31:01 645
en5 English lucifer1004 2021-04-20 11:12:09 2787
en4 English lucifer1004 2021-04-19 06:43:37 82
en3 English lucifer1004 2021-04-19 06:18:41 10
en2 English lucifer1004 2021-04-19 06:09:59 2 (published)
en1 English lucifer1004 2021-04-19 06:06:37 10408 Initial revision (saved to drafts)