Блог пользователя lnzva

Do you often/Are you able to often prove greedy solutions during contests?
Автор lnzva, история, 7 лет назад, По-английски

^

  • Проголосовать: нравится
  • +11
  • Проголосовать: не нравится

»
7 лет назад, # |
  Проголосовать: нравится +80 Проголосовать: не нравится

I usually prove my greedy by submission. :)

→ Ответить
»
7 лет назад, # |
  Проголосовать: нравится +6 Проголосовать: не нравится

There's one argument I use and if it doesn't work, I just believe in it :) Here's it (pretty standard):

Consider candidate to the answer x' ≠ x and where x is the one that would be chosen by greedy. Now we say that if we choose x something gets improved.

→ Ответить
  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится +3 Проголосовать: не нравится

    Proof by contradiction. Can you share some of your favorite greedy problems, I need to improve my problem solving on greedy problems :)

    → Ответить
»
7 лет назад, # |
  Проголосовать: нравится +45 Проголосовать: не нравится

Use the leaderboard. If many people have wrong submissions, it might mean that the obvious-ish greedy solution is wrong. Then one should think about it more. Find a countertest or prove it. Or just stresstest against the naive solution.

I usually think for a moment about a countertest or an easy proof. It doesn't make much sense to spend a lot of time to prove something, unless the solution would take long to implement

→ Ответить
  • »
    »
    7 лет назад, # ^ |
      Проголосовать: нравится +6 Проголосовать: не нравится

    I used the leaderboard in the last contest for the clique problem (Div. 2 E — Div. 1 C Party) and the code passed the pretests. As it turned out, the pretests were just very weak and hundreds of greedy submissions including mine got Wrong answer on test 51 :(

    → Ответить
    • »
      »
      »
      7 лет назад, # ^ |
        Проголосовать: нравится +1 Проголосовать: не нравится

      mehh, just call it a bad day , this trick is mostly godlike , especially to people like me :)

      → Ответить
    • »
      »
      »
      7 лет назад, # ^ |
        Проголосовать: нравится +3 Проголосовать: не нравится

      I was thinking about contests with full feedback, but it's often the same for CF — as long as pretests aren't weak.

      → Ответить
»
7 лет назад, # |
  Проголосовать: нравится +11 Проголосовать: не нравится

I don't have enough experience to advise you but several times I've seen solutions that are solved by some weird comparators then some processing.

Usually I got a little bit surprised, and thought to myself "What sort of thinking process did this guy/team go through", only to ask them later, and know they just implemented some verifier, and tried several comparators till one passed them, submitted and AC.

This is why I usually have a strong dislike against "hard to prove but easy to guess and code" greedy problems.

→ Ответить