結果
問題 | No.5007 Steiner Space Travel |
ユーザー | Forested |
提出日時 | 2022-07-30 17:37:46 |
言語 | C++17 (gcc 12.3.0 + boost 1.83.0) |
結果 |
AC
|
実行時間 | 903 ms / 1,000 ms |
コード長 | 12,530 bytes |
コンパイル時間 | 2,316 ms |
実行使用メモリ | 6,952 KB |
スコア | 8,677,812 |
最終ジャッジ日時 | 2022-07-30 17:38:47 |
合計ジャッジ時間 | 31,396 ms |
ジャッジサーバーID (参考情報) |
judge11 / judge15 |
純コード判定しない問題か言語 |
(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 | AC | 902 ms
4,904 KB |
testcase_01 | AC | 903 ms
6,948 KB |
testcase_02 | AC | 903 ms
4,900 KB |
testcase_03 | AC | 903 ms
6,948 KB |
testcase_04 | AC | 902 ms
6,948 KB |
testcase_05 | AC | 903 ms
4,904 KB |
testcase_06 | AC | 903 ms
4,900 KB |
testcase_07 | AC | 902 ms
6,952 KB |
testcase_08 | AC | 903 ms
4,900 KB |
testcase_09 | AC | 903 ms
4,904 KB |
testcase_10 | AC | 903 ms
6,948 KB |
testcase_11 | AC | 903 ms
4,904 KB |
testcase_12 | AC | 903 ms
4,904 KB |
testcase_13 | AC | 903 ms
6,952 KB |
testcase_14 | AC | 902 ms
4,900 KB |
testcase_15 | AC | 903 ms
4,904 KB |
testcase_16 | AC | 902 ms
6,952 KB |
testcase_17 | AC | 903 ms
6,948 KB |
testcase_18 | AC | 903 ms
4,904 KB |
testcase_19 | AC | 903 ms
4,908 KB |
testcase_20 | AC | 902 ms
4,900 KB |
testcase_21 | AC | 903 ms
4,904 KB |
testcase_22 | AC | 903 ms
4,900 KB |
testcase_23 | AC | 902 ms
4,904 KB |
testcase_24 | AC | 903 ms
4,904 KB |
testcase_25 | AC | 902 ms
4,900 KB |
testcase_26 | AC | 903 ms
4,904 KB |
testcase_27 | AC | 902 ms
4,904 KB |
testcase_28 | AC | 903 ms
6,948 KB |
testcase_29 | AC | 903 ms
4,900 KB |
ソースコード
#ifndef LOCAL #define FAST_IO #endif // ===== template.hpp ===== #include <algorithm> #include <array> #include <bitset> #include <cassert> #include <cmath> #include <iomanip> #include <iostream> #include <list> #include <map> #include <numeric> #include <queue> #include <random> #include <set> #include <stack> #include <string> #include <tuple> #include <unordered_map> #include <unordered_set> #include <utility> #include <vector> #define OVERRIDE(a, b, c, d, ...) d #define REP2(i, n) for (i32 i = 0; i < (i32) (n); ++i) #define REP3(i, m, n) for (i32 i = (i32) (m); i < (i32) (n); ++i) #define REP(...) OVERRIDE(__VA_ARGS__, REP3, REP2)(__VA_ARGS__) #define PER(i, n) for (i32 i = (i32) (n) - 1; i >= 0; --i) #define ALL(x) begin(x), end(x) using namespace std; using u32 = unsigned int; using u64 = unsigned long long; using u128 = __uint128_t; using i32 = signed int; using i64 = signed long long; using i128 = __int128_t; using f64 = double; using f80 = long double; template <typename T> using Vec = vector<T>; template <typename T> bool chmin(T &x, const T &y) { if (x > y) { x = y; return true; } return false; } template <typename T> bool chmax(T &x, const T &y) { if (x < y) { x = y; return true; } return false; } istream &operator>>(istream &is, i128 &x) { i64 v; is >> v; x = v; return is; } ostream &operator<<(ostream &os, i128 x) { os << (i64) x; return os; } istream &operator>>(istream &is, u128 &x) { u64 v; is >> v; x = v; return is; } ostream &operator<<(ostream &os, u128 x) { os << (u64) x; return os; } [[maybe_unused]] constexpr i32 INF = 1000000100; [[maybe_unused]] constexpr i64 INF64 = 3000000000000000100; struct SetUpIO { SetUpIO() { #ifdef FAST_IO ios::sync_with_stdio(false); cin.tie(nullptr); #endif cout << fixed << setprecision(15); } } set_up_io; // ===== template.hpp ===== #ifdef DEBUGF #include "cpl/template/debug.hpp" #else #define DBG(x) (void) 0 #endif // ===== timer.hpp ===== #include <chrono> class Timer { std::chrono::high_resolution_clock::time_point st; public: Timer() : st(std::chrono::high_resolution_clock::now()) {} void reset() { st = std::chrono::high_resolution_clock::now(); } int elapsed() { auto ed = std::chrono::high_resolution_clock::now(); return (int) std::chrono::duration_cast<std::chrono::milliseconds>(ed - st).count(); } }; // ===== timer.hpp ===== // ===== xorshift.hpp ===== class XorShift64 { unsigned long long x; public: XorShift64(unsigned long long seed) : x((seed + 14213124131ull) ^ 103920984124ull) {} unsigned long long operator()() { x = x ^ (x << 13); x = x ^ (x >> 7); x = x ^ (x << 17); return x; } template <typename T> T uniform(T mn, T mx) { return mn + (T) ((*this)() % (mx - mn + 1)); } double as_f64() { return (double) (*this)() / ~0ull; } }; // ===== xorshift.hpp ===== constexpr i32 N = 100; constexpr i32 M = 8; constexpr i32 POS_MIN = 0; constexpr i32 POS_MAX = 1000; [[maybe_unused]] constexpr i32 V_MIN = 1; [[maybe_unused]] constexpr i32 V_MAX = 100000; int dist_two(int x, int y, int p, int q) { return (x - p) * (x - p) + (y - q) * (y - q); } struct Input { i32 n, m; Vec<i32> a, b; }; struct Output { Vec<i32> c, d; i32 v; Vec<i32> t, r; }; Input read_input() { auto &fin = cin; i32 n, m; fin >> n >> m; assert(n == N); assert(m == M); Vec<i32> a(n), b(n); REP(i, n) { fin >> a[i] >> b[i]; assert(POS_MIN <= a[i] && a[i] <= POS_MAX); assert(POS_MIN <= b[i] && b[i] <= POS_MAX); } return Input{n, m, a, b}; } i32 calc_energey(const Input &input, const Output &output, i32 i, i32 j) { i32 x, y, p, q; if (output.t[i] == 0) { x = input.a[output.r[i]]; y = input.b[output.r[i]]; } else { x = output.c[output.r[i]]; y = output.d[output.r[i]]; } if (output.t[j] == 0) { p = input.a[output.r[j]]; q = input.b[output.r[j]]; } else { p = output.c[output.r[j]]; q = output.d[output.r[j]]; } i32 dist = dist_two(p, q, x, y); if (output.t[i] == 0 && output.t[j] == 0) { return 25 * dist; } else if (output.t[i] == 1 && output.t[j] == 1) { return dist; } else { return 5 * dist; } } i64 calc_energey_sum(const Input &input, const Output &output) { i64 s = 0; REP(i, output.v - 1) { s += calc_energey(input, output, i, i + 1); } return s; } pair<Vec<i32>, Vec<i32>> compute_stations_position(const Input &input, XorShift64 &rd) { const auto score = [&input](const Vec<i32> &c, const Vec<i32> &d) -> i64 { i64 sum = 0; REP(i, input.n) { i32 mn = INF; REP(j, input.m) { chmin(mn, dist_two(input.a[i], input.b[i], c[j], d[j])); } sum += 5 * mn; } REP(i, input.m) { i32 mn = INF; REP(j, input.m) { chmin(mn, dist_two(c[i], c[j], c[j], d[j])); } sum += mn; } return sum; }; Vec<i32> c(input.m), d(input.m); REP(i, input.m) { c[i] = rd.uniform(POS_MIN, POS_MAX); d[i] = rd.uniform(POS_MIN, POS_MAX); } i64 cur_score = score(c, d); constexpr i32 TL = 300; Timer timer; i32 iter = 0; while (true) { if (iter % 64 == 0) { if (timer.elapsed() > TL) { break; } } i32 stt = rd.uniform(0, input.m - 1); i32 dx = rd.uniform(-80, 80); i32 dy = rd.uniform(-80, 80); i32 old_x = c[stt], old_y = d[stt]; c[stt] = clamp(c[stt] + dx, POS_MIN, POS_MAX); d[stt] = clamp(d[stt] + dy, POS_MIN, POS_MAX); i64 new_score = score(c, d); i64 diff = new_score - cur_score; if (diff <= 0) { cur_score = new_score; } else { c[stt] = old_x; d[stt] = old_y; } ++iter; } cerr << "iteration: " << iter << '\n'; return make_pair(c, d); } Output solve(Input input) { constexpr i32 TL = 900; Timer timer; XorShift64 rd(998244353); Output output; output.c.resize(input.m, 0); output.d.resize(input.m, 0); output.v = 0; tie(output.c, output.d) = compute_stations_position(input, rd); output.v = input.n + 1; output.t.resize(input.n + 1, 0); output.r.resize(input.n + 1, 0); iota(output.r.begin(), output.r.begin() + input.n, 0); auto uniform_range = [&rd](i32 n) -> pair<i32, i32> { i32 l = rd.uniform(0, n); i32 r = rd.uniform(0, n); while (l == r) { l = rd.uniform(0, n); r = rd.uniform(0, n); } if (l > r) { swap(l, r); } return pair<i32, i32>(l, r); }; auto try_reverse = [&](double temp) -> void { auto [l, r] = uniform_range(output.v - 2); ++l; ++r; i32 diff = 0; diff -= calc_energey(input, output, l - 1, l); diff -= calc_energey(input, output, r - 1, r); diff += calc_energey(input, output, l - 1, r - 1); diff += calc_energey(input, output, l, r); if (diff <= 0 || rd.as_f64() < exp(-diff / temp)) { reverse(output.t.begin() + l, output.t.begin() + r); reverse(output.r.begin() + l, output.r.begin() + r); } }; auto try_insert = [&]([[maybe_unused]] double temp) -> void { i32 stt = rd.uniform(0, input.m - 1); i32 pos = rd.uniform(0, output.v - 2); i32 diff = 0; diff -= calc_energey(input, output, pos, pos + 1); if (output.t[pos] == 0) { diff += 5 * dist_two(input.a[output.r[pos]], input.b[output.r[pos]], output.c[stt], output.d[stt]); } else { diff += dist_two(output.c[output.r[pos]], output.d[output.r[pos]], output.c[stt], output.d[stt]); } if (output.t[pos + 1] == 0) { diff += 5 * dist_two(input.a[output.r[pos + 1]], input.b[output.r[pos + 1]], output.c[stt], output.d[stt]); } else { diff += dist_two(output.c[output.r[pos + 1]], output.d[output.r[pos + 1]], output.c[stt], output.d[stt]); } if (diff < 0 || rd.as_f64() < exp(-diff / temp)) { output.v += 1; output.t.insert(output.t.begin() + pos + 1, 1); output.r.insert(output.r.begin() + pos + 1, stt); } }; auto try_remove = [&]([[maybe_unused]] double temp) -> void { Vec<i32> stt_pos; REP(i, output.v) { if (output.t[i] == 1) { stt_pos.push_back(i); } } if (stt_pos.empty()) { return; } i32 pos = stt_pos[rd.uniform(0, (i32) stt_pos.size() - 1)]; i32 stt = output.r[pos]; i32 diff = 0; diff += calc_energey(input, output, pos - 1, pos + 1); if (output.t[pos] == 0) { diff -= 5 * dist_two(input.a[output.r[pos]], input.b[output.r[pos]], output.c[stt], output.d[stt]); } else { diff -= dist_two(output.c[output.r[pos]], output.d[output.r[pos]], output.c[stt], output.d[stt]); } if (output.t[pos + 1] == 0) { diff -= 5 * dist_two(input.a[output.r[pos + 1]], input.b[output.r[pos + 1]], output.c[stt], output.d[stt]); } else { diff -= dist_two(output.c[output.r[pos + 1]], output.d[output.r[pos + 1]], output.c[stt], output.d[stt]); } if (diff < 0 || rd.as_f64() < exp(-diff / temp)) { output.v -= 1; output.t.erase(output.t.begin() + pos); output.r.erase(output.r.begin() + pos); } }; auto optimize = [&](i32 stt) -> void { Vec<i32> pos; REP(i, output.v) { if (output.t[i] == 1 && output.r[i] == stt) { pos.push_back(i); } } if (pos.empty()) { return; } i64 sq = 0, ln_x = 0, ln_y = 0; for (i32 p : pos) { if (output.t[p - 1] == 0) { sq += 5; ln_x -= 10 * input.a[output.r[p - 1]]; ln_y -= 10 * input.b[output.r[p - 1]]; } else if (output.r[p - 1] != stt) { sq += 1; ln_x -= 2 * output.c[output.r[p - 1]]; ln_y -= 2 * output.d[output.r[p - 1]]; } if (output.t[p + 1] == 0) { sq += 5; ln_x -= 10 * input.a[output.r[p + 1]]; ln_y -= 10 * input.b[output.r[p + 1]]; } else if (output.r[p + 1] != stt) { sq += 1; ln_x -= 2 * output.c[output.r[p + 1]]; ln_y -= 2 * output.d[output.r[p + 1]]; } } output.c[stt] = clamp((i32) round(-ln_x / 2.0 / sq), POS_MIN, POS_MAX); output.d[stt] = clamp((i32) round(-ln_y / 2.0 / sq), POS_MIN, POS_MAX); }; // atode chanto kimeru constexpr double BEGIN_TEMP = 30; constexpr double END_TEMP = 0.1; double temp = BEGIN_TEMP; i32 iter = 0; while (true) { if (iter % 64 == 0) { double t = 1.0 * timer.elapsed() / TL; if (t > 1) { break; } temp = (1 - t) * BEGIN_TEMP + t * END_TEMP; } i32 type = rd.uniform(0, 9); if (type >= 0 && type <= 7) { try_reverse(temp); } else if (type == 8) { try_insert(temp); } else { try_remove(temp); } if (iter % 4096 == 0) { REP(i, input.m) { optimize(i); } } ++iter; } cerr << "iteration: " << iter << '\n'; REP(i, input.m) { optimize(i); } return output; } int main() { Input input = read_input(); Output output = solve(input); REP(i, input.m) { cout << output.c[i] << ' ' << output.d[i] << '\n'; } cout << output.v << '\n'; REP(i, output.v) { cout << output.t[i] + 1 << ' ' << output.r[i] + 1 << '\n'; } //cerr << "score: " << 1e9 / (1000.0 + sqrt(calc_energey_sum(input, output))) << '\n'; }