結果

問題 No.5007 Steiner Space Travel
コンテスト
ユーザー titan23
提出日時 2022-10-06 15:50:44
言語 PyPy3
(7.3.15)
結果
AC  
実行時間 603 ms / 1,000 ms
コード長 3,383 bytes
コンパイル時間 233 ms
実行使用メモリ 85,604 KB
スコア 7,314,610
最終ジャッジ日時 2022-10-06 15:51:06
合計ジャッジ時間 20,510 ms
ジャッジサーバーID
（参考情報） 		judge16 / judge13
純コード判定しない問題か言語
このコードへのチャレンジ
（要ログイン）
ファイルパターン 結果
other AC * 30
権限があれば一括ダウンロードができます

ソースコード

diff # 

import itertools
import random
import time
start = time.time()
N, M = map(int, input().split())
AB = [list(map(int, input().split())) for _ in range(N)]

class KMeans:

  def __init__(self, n_clusters, max_iter = 10, random_seed = 0):
    self.n_clusters = n_clusters
    self.max_iter = max_iter

  def fit(self, X):
    self.labels_ = []
    now = 0
    while len(self.labels_) < len(X):
      self.labels_.append(now)
      now += 1
      now %= self.n_clusters
    random.shuffle(self.labels_)
    labels_prev = [0]*len(X)
    count = 0
    self.cluster_centers_ = [(0, 0)] * self.n_clusters

    while (not (self.labels_ == labels_prev) and count < self.max_iter):
      syuukei = [[] for _ in range(self.n_clusters)]
      for i in range(len(X)):
        syuukei[self.labels_[i]].append(X[i])
      for i,l in enumerate(syuukei):
        if l:
          x, y = sum(x for x,y in l)//len(l), sum(y for x,y in l)//len(l)
        else:
          x, y = random.randint(0, 1000), random.randint(0, 1000)
        self.cluster_centers_[i] = (x, y)

      labels_prev = self.labels_[:]
      for i in range(len(X)):
        dist = 10**18
        for j in range(self.n_clusters):
          tmp = (X[i][0] - self.cluster_centers_[j][0])**2 + (X[i][1] - self.cluster_centers_[j][1])**2
          if tmp < dist:
            dist = tmp
            self.labels_[i] = j
      count += 1


def main():
  model = KMeans(8)
  model.fit(AB)
  labels = model.labels_
  centers = model.cluster_centers_

  def dist(a, b):
    return (a[0] - b[0])**2 + (a[1] - b[1])**2

  def calc(ans):
    score = 0
    for i in range(len(ans)-1):
      type_pre, indx_pre = ans[i]
      type_now, indx_now = ans[i+1]
      if type_pre == 1 and type_now == 1:
        score += 25*dist(AB[indx_pre], AB[indx_now])
      if type_pre == 1 and type_now == 2:
        score +=  5*dist(AB[indx_pre], centers[indx_now])
      if type_pre == 2 and type_now == 1:
        score +=  5*dist(centers[indx_pre], AB[indx_now])
      if type_pre == 2 and type_now == 2:
        score +=  1*dist(centers[indx_pre], centers[indx_now])
    return score

  ANS = []
  L = [[] for _ in range(8)]
  for i, cluster_number in enumerate(labels):
    L[cluster_number].append(i)

  def calc_p(p):
    score = dist((AB[0]), centers[p[0]])
    for i in range(len(centers)-1):
      score += dist(centers[p[i]], centers[p[i+1]])
    score += dist(centers[p[-1]], AB[0])
    return score

  vest_p = None
  vest = 1<<30
  for p in itertools.permutations(range(8)):
    score = calc_p(p)
    if score < vest:
      vest = score
      vest_p = p[:]
  if time.time() - start > 0.9:
    return 1<<30, -1, -1
  ans = [(1, 0)]
  for i in vest_p:
    if i >= len(L): continue
    ans.append((2, i))
    lim = len(L[i])
    j = 0
    while j < lim:
      d1 = 5 * dist(AB[L[i][j]], centers[i])
      d2 = 1<<30
      if j+1 < len(L[i]):
        d2 = 25* dist(AB[L[i][j]], AB[L[i][j+1]])
      if d2 < d1:
        ans.append((1, L[i][j]))
      else:
        ans.append((1, L[i][j]))
        ans.append((2, i))
      j += 1
  ans.append((1, 0))
  score = calc(ans)
  return score, ans, centers

ANS = []
CENTERS = []
vest = 1<<30
while time.time() - start < 0.5:
  tmp, ans, centers = main()
  if tmp < vest:
      vest = tmp
      ANS = ans[:]
      CENTERS = centers

for c,d in CENTERS:
  print(c, d)
print(len(ANS))
for a,b in ANS:
  print(a, b+1)
0