結果

問題 No.5020 Averaging
コンテスト
ユーザー uuuus17
提出日時 2024-02-25 15:55:34
言語 PyPy3
(7.3.15)
結果
AC  
実行時間 906 ms / 1,000 ms
コード長 5,140 bytes
コンパイル時間 372 ms
コンパイル使用メモリ 81,828 KB
実行使用メモリ 290,840 KB
スコア 22,799,922
最終ジャッジ日時 2024-02-25 15:56:27
合計ジャッジ時間 46,764 ms
ジャッジサーバーID
（参考情報） 		judge15 / 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 * 50
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 sys
input = lambda: sys.stdin.readline().rstrip()
# sys.setrecursionlimit(10**7)
# sys.set_int_max_str_digits(10**6)
# import pypyjit
# pypyjit.set_param('max_unroll_recursion=-1')
def mp():return map(int,input().split())
def lmp():return list(map(int,input().split()))
def lm1(LIST): return list(map(lambda x:x-1, LIST))
def mps(A):return [tuple(map(int, input().split())) for _ in range(A)]
def stoi(LIST):return list(map(int,LIST))
def itos(LIST):return list(map(str,LIST))
def atoi(LIST): return [ord(i)-ord("a") for i in LIST]
def Atoi(LIST): return [ord(i)-ord("A") for i in LIST]
def LT(LIST,N): return LIST[bisect.bisect_left(LIST,N)-1]
def LE(LIST,N): return LIST[bisect.bisect_right(LIST,N)-1]
def GT(LIST,N): return LIST[bisect.bisect_right(LIST,N)]
def GE(LIST,N): return LIST[bisect.bisect_left(LIST,N)]
def bitA(X,A):return X & 1<<A == 1<<A
def gtoi(x,y,h,w):return x*w+y
import math
import bisect
import heapq
import time
import random as rd
import itertools
from copy import copy as cc
from copy import deepcopy as dc
from itertools import accumulate, product
from collections import Counter, defaultdict, deque
def ceil(U,V):return (U+V-1)//V
def modf1(N,MOD):return (N-1)%MOD+1
def printmat(list):
    for i in list:print(*i)
m4 = [[1,0],[0,1],[-1,0],[0,-1]]
m8 = [[-1,-1],[-1,0],[-1,1],[0,-1],[0,1],[1,-1],[1,0],[1,1]]
inf = (1<<63)-1
mod = 998244353

start_time = time.time()
n = int(input())
omote = []
ura = []
max_limit = 50
for i in range(n):
    a,b = mp()
    omote.append(a)
    ura.append(b)
width = 5
depth = 3

def output(sosa_arr):
    print(len(sosa_arr))
    for i,j in sosa_arr:print(i+1,j+1)

def calc_score(x,y):
    return max(abs(500000000000000000-x),abs(500000000000000000-y))

def calc_all_score(a,b,X):
    ErrorA = abs(500000000000000000 - a[0])
    ErrorB = abs(500000000000000000 - b[0])
    Score = (int)(2000000.0 - 100000.0 * math.log10(1.0 * max(ErrorA, ErrorB) + 1.0))
    if ErrorA == 0 and ErrorB == 0:
        Score = 2000050 - X
    return Score


def greedy(a,b):
    sosa = []
    for kaisu in range(max_limit):
        score = calc_score(a[0],b[0])
        choose_card = 0
        for i in range(1,n):
            na,nb = (a[0]+a[i])/2, (b[0]+b[i])/2
            now_score = calc_score(na,nb)
            if now_score < score:
                score = now_score
                choose_card = i
        a[0],b[0],a[choose_card],b[choose_card] = (a[0]+a[choose_card])/2, (b[0]+b[choose_card])/2,\
                                                  (a[0]+a[choose_card])/2, (b[0]+b[choose_card])/2
        if choose_card == 0:break
        sosa.append([0,choose_card])
    return a,b,sosa


def remove_huyou(A):
    while len(A) != 1:
        if A[-1] == A[-2]:A.pop()
        else:break
    return A


def hanei_0(Score,A,B,Sosa):
    choose_card = 0
    Score = calc_score(A[0],B[0])
    for i in range(1, n):
        na, nb = (A[0] + A[i]) / 2, (B[0] + B[i]) / 2
        now_score = calc_score(na, nb)
        if now_score < Score:
            Score = now_score
            choose_card = i
    A[0], B[0], A[choose_card], B[choose_card] = (A[0] + A[choose_card]) / 2, (B[0] + B[choose_card]) / 2, \
                                                 (A[0] + A[choose_card]) / 2, (B[0] + B[choose_card]) / 2
    if choose_card != 0:Sosa.append([0, choose_card])
    return [Score,A,B,Sosa]


def beam_search(a,b,sosa_Arr):
    koho = [[calc_score(a[0],b[0]), cc(a), cc(b), sosa_Arr]]  # 次stepで反映したときのscore,今のa,今のb,今の操作
    heapq.heapify(koho)
    best_score = calc_score(a[0],b[0])
    for dep in range(depth):
        now_koho = []
        while not (len(koho) == 0 or len(now_koho) == width):
            nxt = heapq.heappop(koho)
            if nxt[0] > best_score:break
            now_koho.append(nxt)
        koho.clear()
        #print("n",now_koho)
        best_score = now_koho[0][0]
        for Score,A,B,Sosa in now_koho:
            if len(Sosa) == max_limit:continue
            heapq.heappush(koho, dc([Score, A, B, Sosa]))
            for i in range(n):
                for j in range(i+1,n):
                    if Sosa and Sosa[-1] == [i,j]:continue
                    ac = cc(A)
                    bc = cc(B)
                    sosac = cc(Sosa)
                    na,nb = (a[i]+a[j])/2, (b[i]+b[j])/2
                    ac[i],ac[j],bc[i],bc[j] = na,na,nb,nb
                    hanei_a,hanei_b = (na+ac[0])/2, (nb+bc[0])/2
                    now_score = calc_score(hanei_a,hanei_b)
                    sosac.append([i,j])
                    heapq.heappush(koho,[now_score,ac,bc,sosac])
        #print(koho)
    best_koho = heapq.heappop(koho)
    best_koho[3] = remove_huyou(best_koho[3])
    best_koho = hanei_0(*best_koho)
    return best_koho


def main():
    sosa_arr = []
    global omote
    global ura
    while time.time()-start_time <= 0.8:
        pred_score,na,nb,sosa_arr = beam_search(omote,ura,sosa_arr)
        omote = na
        ura = nb
        #print(pred_score,omote,ura,sosa_arr)
    output(sosa_arr)
    #print(calc_all_score(omote,ura,len(sosa_arr)))
    #na,nb,sosa_arr = greedy(omote,ura)

main()
0