結果

問題 No.5020 Averaging
ユーザー A8pfA8pf
提出日時 2024-02-25 15:55:25
言語 C++17
(gcc 12.3.0 + boost 1.83.0)
結果
WA  
実行時間 -
コード長 5,837 bytes
コンパイル時間 3,159 ms
コンパイル使用メモリ 234,264 KB
実行使用メモリ 6,676 KB
スコア 0
最終ジャッジ日時 2024-02-25 15:55:41
合計ジャッジ時間 15,873 ms
ジャッジサーバーID
（参考情報） 		judge11 / judge15
このコードへのチャレンジ
（要ログイン）

テストケース

テストケース表示
入力 結果 実行時間
実行使用メモリ
testcase_00 WA -
testcase_01 WA -
testcase_02 WA -
testcase_03 WA -
testcase_04 WA -
testcase_05 WA -
testcase_06 WA -
testcase_07 WA -
testcase_08 WA -
testcase_09 WA -
testcase_10 WA -
testcase_11 WA -
testcase_12 WA -
testcase_13 WA -
testcase_14 WA -
testcase_15 WA -
testcase_16 WA -
testcase_17 WA -
testcase_18 WA -
testcase_19 WA -
testcase_20 WA -
testcase_21 WA -
testcase_22 WA -
testcase_23 WA -
testcase_24 WA -
testcase_25 WA -
testcase_26 WA -
testcase_27 WA -
testcase_28 WA -
testcase_29 WA -
testcase_30 WA -
testcase_31 WA -
testcase_32 WA -
testcase_33 WA -
testcase_34 WA -
testcase_35 WA -
testcase_36 WA -
testcase_37 WA -
testcase_38 WA -
testcase_39 WA -
testcase_40 WA -
testcase_41 WA -
testcase_42 WA -
testcase_43 WA -
testcase_44 WA -
testcase_45 WA -
testcase_46 WA -
testcase_47 WA -
testcase_48 WA -
testcase_49 WA -
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 # 

#include<bits/stdc++.h>
using namespace std;
using ull=unsigned long long;

std::mt19937 mt_for_action(0); // 行動選択用の乱数生成器を初期化

// 128bit整数
std::ostream &operator<<(std::ostream &dest, __int128_t value) {
  std::ostream::sentry s(dest);
  if (s) {
    __uint128_t tmp = value < 0 ? -value : value;
    char buffer[128];
    char *d = std::end(buffer);
    do {
      --d;
      *d = "0123456789"[tmp % 10];
      tmp /= 10;
    } while (tmp != 0);
    if (value < 0) {
      --d;
      *d = '-';
    }
    int len = std::end(buffer) - d;
    if (dest.rdbuf()->sputn(d, len) != len) {
      dest.setstate(std::ios_base::badbit);
    }
  }
  return dest;
}

__int128 parse(string &s) {
  __int128 ret = 0;
  for (int i = 0; i < s.length(); i++)
    if ('0' <= s[i] && s[i] <= '9')
      ret = 10 * ret + s[i] - '0';
  return ret;
}

// 128bit整数ここまで
constexpr int MAX_TURN = 50;
constexpr __int128 TARGET = 500000000000000000;

struct Card {
  __int128 omote, ura;
};

class GameState
{
  private:
  int _turn;
  vector<Card> _cards;
  vector<pair<int, int>> _commandHistory;

  public:
  int nowScore;
  pair<int, int> first_action_;
  GameState(){};
  GameState(vector<unsigned long long> omotes, vector<unsigned long long> uras): _turn(0), nowScore(0) {
    for (int i = 0; i < omotes.size(); i++) {
      _cards.push_back({omotes[i], uras[i]});
    }
  }

  __int128 getEstimatedScore() {
    __int128 score = 0;
    // 1枚目のみ別に評価
    __int128 omotediff = TARGET - _cards[0].omote;
    if (omotediff > 0) {
      omotediff *= -1;
    }
    score += omotediff;
    __int128 uradiff = TARGET - _cards[0].ura;
    if (uradiff > 0) {
      uradiff *= -1;
    }
    score += uradiff;
    // 満点ならその時点で終了
    if (omotediff == 0 && uradiff == 0) {
      score += 9999999999999999;
    }
    // 2枚目以降
    const Card OTHER_TARGET = { TARGET * 2 - _cards[0].omote, TARGET * 2 - _cards[0].ura };
    for (int i = 1; i < _cards.size(); i++) {
      __int128 omotediff = OTHER_TARGET.omote - _cards[i].omote;
      if (omotediff > 0) {
        omotediff *= -1;
      }
      score += omotediff * _turn;
      __int128 uradiff = OTHER_TARGET.ura - _cards[i].ura;
      if (uradiff > 0) {
        uradiff *= -1;
      }
      score += uradiff * _turn;
      // 満点ならその時点で終了
      if (omotediff == 0 && uradiff == 0) {
        score += 99999999;
      }
    }
    return score;
  }

  // 1ターン進める
  void advance(int u, int v)
  {
    _commandHistory.push_back({u, v});
    _turn++;
    Card newCard = {(_cards[u].omote + _cards[v].omote) / 2, (_cards[u].ura + _cards[v].ura) / 2};
    _cards[u] = newCard;
    _cards[v] = newCard;
    nowScore = getEstimatedScore();
  }

  // 合法手を列挙する
  vector<pair<int, int>> legalActions() const
  {
    vector<pair<int, int>> ret;
    for (int i = 1; i < _cards.size(); i++) {
      ret.push_back({0, i});
    }
    return ret;
  }
  
  // 1番目に触らないランダムなアクションを返す
  pair<int, int> randomAction() const
  {
    uniform_int_distribution<int> dist(1, _cards.size());
    int u = dist(mt_for_action);
    int v = dist(mt_for_action);
    while (u == v) {
      v = dist(mt_for_action);
    }
    return {u, v};
  }

  // ゲームが終了しているかどうか
  bool isDone() const
  {
    return _turn == MAX_TURN;
  }

  // 結果の出力を行う
  string outputCommand() {
    stringstream ss;
    ss << _commandHistory.size() << endl;
    for (auto command : _commandHistory) {
      ss << command.first + 1 << " " << command.second + 1 << endl;
    }
    return ss.str();
  }

  // 現在の実スコアを出力する
  double getScore() const {
    ull v = max( max((ull)_cards[0].omote, (ull)TARGET) - min((ull)_cards[0].omote, (ull)TARGET), max((ull)_cards[0].ura, (ull)TARGET) - min((ull)_cards[0].ura, (ull)TARGET) );
    long long score = 2000000 - 100000 * log10(static_cast<long double>(v));
    return score;
  }
};

bool operator<(const GameState &state1, const GameState &state2)
{
  return state1.nowScore < state2.nowScore;
}

pair<int,int> beamSearchAction(const GameState &initState, const int beamWidth, const int beamDepth)
{
  priority_queue<GameState> nowBeam;
  GameState bestState;

  nowBeam.push(initState);
  for (int t = 0; t < beamDepth; t++)
  {
      std::priority_queue<GameState> nextBeam;
      for (int i = 0; i < beamWidth; i++)
      {
          if (nowBeam.empty())
              break;
          GameState nowState = nowBeam.top();
          nowBeam.pop();
          auto legal_actions = nowState.legalActions();
          for (const auto &action : legal_actions)
          {
              GameState nextState = nowState;
              nextState.advance(action.first, action.second);
              if (t == 0)
                  nextState.first_action_ = action;
              nextBeam.push(nextState);
          }
      }
      nowBeam = nextBeam;
      bestState = nowBeam.top();
      if (bestState.isDone())
      {
        break;
      }
  }
  if (bestState.getScore() < initState.getScore()) {
    bestState = initState;
    auto random_action = initState.randomAction();
    return random_action;
  }
  return bestState.first_action_;
}

pair<vector<unsigned long long>, vector<unsigned long long>> getFromInput() {
  int n;
  cin >> n;
  vector<unsigned long long> omotes(n), uras(n);
  for (int i = 0; i < n; i++) {
    cin >> omotes[i] >> uras[i];
  }
  return {omotes, uras};
}

int main() {
  // 本番用
  auto input = getFromInput();
  GameState state(input.first, input.second);
  while (!state.isDone()) {
    auto action = beamSearchAction(state, 10, 10);
    state.advance(action.first, action.second);
    // cerr<<state.getScore()<<endl;
  }
  cout << state.outputCommand();
}
0