SuprDewd's blog

By SuprDewd, history, 4 years ago, In English

NWERC 2020 was finally held this last weekend after a few month delay due to the pandemic. The format was a little different this time around, with each contestant participating from home (i.e. three computers per team) and having full read access to the internet during the contest.

We wanted to let you know that this Saturday at 12:00 GMT we will be hosting an open version of the contest on Kattis: https://open.kattis.com/contests/nwerc2020open

The problemset, solution slides, testdata and judges' solutions are already available on our website, but we ask that you please refrain from looking at them until the open contest is over. Good luck, and we hope you enjoy the problems!

  • Vote: I like it
  • +35
  • Vote: I do not like it

»
4 years ago, # |
  Vote: I like it 0 Vote: I do not like it

This coincides with the div 1 codeforces round :(

  • »
    »
    4 years ago, # ^ |
      Vote: I like it 0 Vote: I do not like it

    That's unfortunate :( I don't think we can change the time now, sadly.

»
4 years ago, # |
  Vote: I like it +8 Vote: I do not like it

We had a solution to B that is very different from the jury's solution.

We define the function $$$f(i, d)$$$, which is given by

  • If $$$d \geq 0$$$, the number of operations we need to level everything, if positions $$$i, i+1, \dots, n$$$ are as normal, positions $$$i-1, \dots, i-d$$$ are $$$0$$$, and positions $$$i-d-1, i-d-2, \dots$$$ are $$$1$$$.
  • If $$$d < 0$$$, the number of operations we need to level everything SUBTRACTED BY $$$|d|$$$, if positions $$$i, i+1, \dots, n$$$ are as normal, positions $$$i-2, i-3, \dots$$$ are $$$1$$$, and position $$$i-1$$$ is $$$1-d$$$.

The answer is given by $$$f(0, \infty)$$$, and $$$f(n, d) = 0$$$ for all $$$d$$$.

Now, if $$$a[i] = 0$$$, then $$$f(i, d) = f(i+1, d+1)$$$.

Otherwise,

  • Let $$$0 < j_1 \leq \infty$$$ be the minimum value s.t. $$$f(i+1, j_1) < f(i+1, j_1 - 1)$$$ (this is the first amount where we want to push blocks past the left edge).
  • Let $$$-\infty \leq j_2 <= 0$$$ be the maximum value s.t. $$$f(i+1, j_2) = f(i+1, j_2 - 1)$$$ (this is the first amount where we want to push the extra blocks past the right edge).

Then, $$$f(i, d) = f(i+1, d - (a[i] - 1)) + (a[i] - 1)$$$ + $$$a(i, d)$$$, where $$$a(i, d) = 0$$$ if $$$d \in [j_2 + (a[i] - 1), j_1 + (a[i] - 1) - 1]$$$, and $$$1$$$ otherwise.

These updates can be made in $$$\mathcal{O}(\log n)$$$, though our implementation was $$$\mathcal{O}(\log^2 n)$$$. Thus, total complexity is $$$\mathcal{O}(n \log n)$$$.

code
»
4 years ago, # |
  Vote: I like it +21 Vote: I do not like it

The problems are available for upsolving on Kattis, but by popular demand we have also uploaded the contest to the Codeforces gym.

»
3 years ago, # |
  Vote: I like it +37 Vote: I do not like it

In problem I, my teammate AsunderSquall used a weird algorithm and magically got accepted.

There is a simple $$$O(n^4)$$$ brute force method and it's too slow. But we add this line :

if (clock() * 1.0 / CLOCKS_PER_SEC > 6.9) { puts("impossible"); exit(0); }

It means if we can't find a vaild solution in TL, the program just gives up. Then we passed the test.


After the (VP)contest we guessed that if the solution exists, we can just make the first person start at $$$1$$$.

Can anyone proof this, or figure out it is fake and passed just because of weak test?