Tips for heuristic/bot programming contests
- masashinakata
- 781
- 0
- 0
- 0
#82: Chasing current metagame and imitating top players might give you good placement, but won't teach you too much in the long-term. If your goal is self-improvement, try avoiding those approaches.
2022-12-23 09:45:01I think I have touched all of the aspects that I wanted to. Obviously, those takes are shallow compared to how much wisdom there is to share, but I'll never be able to do that with a twitter thread. I'll leave the rest for Q&A (pls ask questions!) and maybe wrap-up. Going 💤
2022-12-23 09:52:46@FakePsyho Thanks for the great Tips! There are so many competitors in Japan who look up to you. This series of tweets will have a significant impact on Japanese competitors in the fields of heuristics and bots.
2022-12-23 10:15:50@ozy4dm これですか? twitter.com/tomerun/status…
2022-12-23 14:20:29はてなブログに投稿しました #はてなブログ BFSを繰り返すときに訪問済みかを記録する配列を毎回初期化しなくて良くするアレ - TopCoderの学習のお時間 topcoder-tomerun.hatenablog.jp/entry/2022/11/…
2022-11-06 14:52:08@tomerun あーこれです!ありがとうございます。Psyhoさんのツイートに同じようなことが書いてあったので、併せて書いておこうかと思いまして。
2022-12-23 14:22:03残りのツイートも全部まとめました! 見出しも付けたので読みやすくなったはず!! はてなブログに投稿しました #はてなブログ Psyhoさんによるヒューリスティック・ボットコンテストのための無料Tips - カメヲラボ ozy4dm.hateblo.jp/entry/2022/12/…
2022-12-23 16:51:33@FakePsyho "Fast-Clearing Array replaces sets/maps" means using array indices as hash keys?
2022-12-23 20:07:05@ToastUz Yes! Common example would be avoiding duplicated states in beam search.
2022-12-23 20:23:29ヒューリスティックに関してはまだまだひよっ子だという自覚はあるけど、それでも間違いなく大事だと思うのは「暇だろうが忙しかろうが、やる気があろうがなかろうが、とりあえず参加登録はしておく」ってことかな。
2022-12-24 07:41:04@mshiladityam Surprisingly it wasn't a typo. I do recall exactly a single case when someone was able to use advanced math to achieve a better result. Very old and very unique marathon.
2022-12-24 18:13:31ほうほうと思って読んでみたら「Topcoderにはdimkadimon、AtCoderにはwataがいて素晴らしい仕事をしている」みたいな、めたんこカッコ良いことが書かれていた🤩 twitter.com/tomerun/status…
2022-12-24 20:12:16codeforces.com/blog/entry/110… Mikeさんがcodeforcesで定期的なヒューリスティックコンテストを開催することに関心を示している
2022-12-24 18:58:54@FakePsyho Regarding reallocating memory every call. I've heard other people also believing that reusing global vector<> and clearing gives better speed, but from my experience, it's often the contrary. I think the reason is compiler can optimize much better when vector is a local variable
2022-12-25 17:12:45@cdkrot Unsure what's the exact comparison: global vector cleared (= no de-allocation) vs local vector allocated every time? If the latter isn't slower by at least 5-10x that would go against everything I saw so far. Also, in theory simple arrays should be slightly faster than vectors
2022-12-25 19:43:10Probably 1-2 more days and I'll make it public. Needs some proper docs, better error handling and some basic stress tests to make sure that file communication is fine. If you have a program that is able to run 2 AIs and output result, setup should take less than 2 minutes. twitter.com/FakePsyho/stat…
2022-12-28 00:01:32Not really a tip, but I've been working on a simple local league system that would do automatic matchmaking. Somewhat similar tool to mmtester but for bot contests. Hopefully in the near future this will be published as a PyPI package.
2022-12-23 09:38:18@FakePsyho Nice ! If you have any "beta" version to spare, I'd be glad to test it :)
2022-12-28 00:42:11マラソン系コンテストの環境構築記事のメモ AWS(Lambda):yunix-kyopro.hatenablog.com/entry/2022/07/… GCP: threecourse.hatenablog.com/entry/2020/07/… まとめ系: kmyk.github.io/blog/blog/2021… AWS(EC2): takoshi.hatenablog.com/entry/2018/12/…
2022-12-28 20:31:54