How am I

26 year's old
I graduated from Graduate School of Engineering, Information and Systems, University of Tsukuba.
I'm working at ACCESS CO.,LTD.

What am I interested in

Parallel computing
3D Display
Machine learning

What do I

I'm developing Shogi (Japanese chess) AI, ponkotsu.
I'm developing reversi AI, dekunobou using genetic algorithm.
I enjoy competitive programming. I'm green coder (Highest 1092) at AtCoder.

Affiliation

Achievements

2019/08/03, 04: Exhibit Maker Faire Tokyo 2019 as a menber of OpenEsys. (link)
2020/03/08: Got 800 rating and became green coder at AtCoder.
2021/05/03: Participated World Computer Shogi Championship (WCSC31), and got 28th prize at first qualifyng round.(2 wins 6 lose)
2021/06/27: Got 800 rating and became green coder at AtCoder Heuristic Contest(AHC)
2021/07/03: Participated Denryusen TSEC2 and got 40th prize at prelium round.(10 wins 13 lose 1 draw)
2021/11/20, 21: Participated Denryusen and got 15th prize at C-class round.(4 wins 13 lose)
2022/05/03-05: Participated World Computer Shogi Championship (WCSC32), and got 19th prize at first qualifyng round.(4 wins 4 lose)
2023/05/03-05: Participated World Computer Shogi Championship (WCSC33), and got 11th prize.(11 wins 6 lose)
202３/12/02-03: Participated Denryusen and got 1st place at C-class round.(12 wins 1 lose 1 draw)
2024/01/12~03/17: Participated Denryusen Mynavi News Cup and got 10th prize at qualifying round.(13 wins 14 lose 3 draw)
2024/05/03-05/05: Participated World Computer Shogi Championship (WCSC34), and got 13th prize.(3 wins 2 lose 4 draw)

My skills

C/C++
Python3
Fortran
OpenGL
Pytorch
libtorch
OpenMP

Studying skills

Rust
Go
Ruby on Rails

Contact

Misskey: @[email protected]
mail: jj1guj[AT]jj1guj.net (Please change [AT] to @)

adaptation_systems's People

Contributors

Stargazers

Watchers

adaptation_systems's Issues

judge_task1.pyにおけるジャッジの並列化

今回TLが10分以内なので50個のテストケースに対してシングルスレッドでジャッジを行うとジャッジが完了するまでに8時間以上かかってしまう.
そのため、より円滑にジャッジを終了させるために並列化したいです.

judgeコードの浮動小数桁落ち/丸め誤差について

ジャッジコードでは浮動小数を用いてスコア計算をしていますが、これだと丸め誤差と桁落ちで正確な点数が計算できないと思います。
PythonなのでfloatをDecimalで置き換えるだけで良さそうです(私はあまりDecimalに詳しくありませんが…)
もしくは多倍長整数の恩恵を享受して、小数点を抜いた整数として値を扱っても良いかもしれないと思いました。

make_datasetの入力の最大桁数について

入力の最大桁数は30桁なので、浮動小数を使って入力を作るのではなく、30桁のそれぞれについてrandintを使って桁を生成する方が良いと思います。

私は手元で以下のように改造しました(コードの挿入方法がわからずベタ書きですみません)

def reduce_digit(s):
for i in reversed(range(len(s))):
if s[i] == '0':
s = s[:i]
else:
break
return s

for d in range(datasets):
N=random.randint(min(args.data_min,args.data_max),max(args.data_min,args.data_max))
A=[str('0.' + ''.join(str(random.randint(0, 9)) for _ in range(30))) for i in range(N)]
A=[reduce_digit(i) for i in A]
filename=os.path.join(path,str(d+args.seed).zfill(4)+".in")
with open(filename,mode='w') as f:
f.write("\n".join(A))