use std::cmp::Reverse; use std::collections::{HashMap, HashSet}; use std::io::Write; use std::{collections::BinaryHeap, fmt::Display}; use grid::{inv, Coordinate, Map2d, ADJACENTS}; use judge::{OnlineJudge, SelfJudge}; use crate::{judge::Judge, rand::Xoshiro256}; pub trait ChangeMinMax { fn change_min(&mut self, v: Self) -> bool; fn change_max(&mut self, v: Self) -> bool; } impl ChangeMinMax for T { fn change_min(&mut self, v: T) -> bool { *self > v && { *self = v; true } } fn change_max(&mut self, v: T) -> bool { *self < v && { *self = v; true } } } macro_rules! get { ($t:ty) => { { let mut line: String = String::new(); std::io::stdin().read_line(&mut line).unwrap(); line.trim().parse::<$t>().unwrap() } }; ($($t:ty),*) => { { let mut line: String = String::new(); std::io::stdin().read_line(&mut line).unwrap(); let mut iter = line.split_whitespace(); ( $(iter.next().unwrap().parse::<$t>().unwrap(),)* ) } }; ($t:ty; $n:expr) => { (0..$n).map(|_| get!($t) ).collect::>() }; ($($t:ty),*; $n:expr) => { (0..$n).map(|_| get!($($t),*) ).collect::>() }; ($t:ty ;;) => { { let mut line: String = String::new(); std::io::stdin().read_line(&mut line).unwrap(); line.split_whitespace() .map(|t| t.parse::<$t>().unwrap()) .collect::>() } }; ($t:ty ;; $n:expr) => { (0..$n).map(|_| get!($t ;;)).collect::>() }; } #[allow(unused_macros)] macro_rules! chmin { ($base:expr, $($cmps:expr),+ $(,)*) => {{ let cmp_min = min!($($cmps),+); if $base > cmp_min { $base = cmp_min; true } else { false } }}; } #[allow(unused_macros)] macro_rules! chmax { ($base:expr, $($cmps:expr),+ $(,)*) => {{ let cmp_max = max!($($cmps),+); if $base < cmp_max { $base = cmp_max; true } else { false } }}; } #[allow(unused_macros)] macro_rules! min { ($a:expr $(,)*) => {{ $a }}; ($a:expr, $b:expr $(,)*) => {{ std::cmp::min($a, $b) }}; ($a:expr, $($rest:expr),+ $(,)*) => {{ std::cmp::min($a, min!($($rest),+)) }}; } #[allow(unused_macros)] macro_rules! max { ($a:expr $(,)*) => {{ $a }}; ($a:expr, $b:expr $(,)*) => {{ std::cmp::max($a, $b) }}; ($a:expr, $($rest:expr),+ $(,)*) => {{ std::cmp::max($a, max!($($rest),+)) }}; } const N: usize = 14; #[derive(Debug, Clone)] struct Input { n: usize, t: usize, people: Vec, sampled_people: Vec, sampled_houses: HashSet, used_count_dict: Vec, } impl Input { fn read_input() -> Input { let (n, t) = get!(usize, usize); let mut people = vec![]; let mut counts = Map2d::new(vec![0; N * N], N); for _ in 0..n { let (r0, c0, r1, c1) = get!(usize, usize, usize, usize); let mut c0 = Coordinate::new(r0 - 1, c0 - 1); let mut c1 = Coordinate::new(r1 - 1, c1 - 1); if c0 > c1 { std::mem::swap(&mut c0, &mut c1); } people.push(Person::new(c0, c1)); counts[c0] += 1; counts[c1] += 1; } let mut candidates = vec![]; for row in 0..N { for col in 0..N { let c = Coordinate::new(row, col); candidates.push(c); } } const TARGET_COUNT: usize = 30; candidates.sort_unstable_by_key(|c| Reverse(counts[*c])); candidates.truncate(TARGET_COUNT); let mut sampled_people = vec![]; let mut sampled_houses = HashSet::new(); for &c in &candidates { sampled_houses.insert(c); } for &person in &people { if sampled_houses.contains(&person.home) { sampled_people.push(person); } else if sampled_houses.contains(&person.company) { sampled_people.push(Person::new(person.company, person.home)); } } let mut used_count_dict = vec![0; 30000]; for not_used in 0..30 { for used in 0..30 { let cost = 1000 * not_used + 223 * used; if cost >= used_count_dict.len() { break; } used_count_dict[cost] = used as i64; } } Input { n, t, people, sampled_people, sampled_houses, used_count_dict, } } } #[derive(Debug, Clone, Copy)] struct Person { home: Coordinate, company: Coordinate, } impl Person { fn new(home: Coordinate, company: Coordinate) -> Self { Self { home, company } } } #[derive(Debug, Clone)] struct State { money: i64, collaborator: i64, map: Map2d<[bool; 4]>, } impl State { fn init() -> Self { let map = Map2d::new(vec![[false; 4]; N * N], N); Self { money: 1000000, collaborator: 1, map, } } fn calc_construction_cost(&self) -> i64 { for i in 1..100 { if i * i == self.collaborator { return 10_000_000 / i; } } (1e7 / (self.collaborator as f64).sqrt()).floor() as i64 } fn can_construct(&self) -> bool { self.money >= self.calc_construction_cost() } } enum Action { Construct(Coordinate, Coordinate), Collaboration, Money, } impl Action { fn apply(&self, state: &mut State) { if let &Action::Construct(p, q) = self { let mut dir = !0; for d in 0..4 { let adj = ADJACENTS[d]; let next = p + adj; if next == q { dir = d; break; } } state.map[p][dir] = true; state.map[q][inv(dir)] = true; } } } impl Display for Action { fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result { match self { Action::Construct(p, q) => write!( f, "1 {} {} {} {}", p.row + 1, p.col + 1, q.row + 1, q.col + 1 ), Action::Collaboration => write!(f, "2"), Action::Money => write!(f, "3"), } } } fn main() { let mut state = State::init(); let mut judge = get_judge(); let input = judge.read_input(); let mut blueprint = annealing(&input, AnnealingState::new(), 1.85); blueprint.update_order(); for turn in 0..input.t { judge.update_state(&mut state); let action = get_action(&input, &state, &blueprint, turn); action.apply(&mut state); judge.apply(action); } } fn get_judge() -> Box { match std::env::var("LOCAL") { Ok(_) => Box::new(SelfJudge::new()), Err(_) => Box::new(OnlineJudge), } } fn get_action(input: &Input, state: &State, blueprint: &AnnealingState, turn: usize) -> Action { if turn < 80 { return Action::Collaboration; } if turn >= 350 || !state.can_construct() { return Action::Money; } for &(p, q) in &blueprint.candidates { let mut dir = !0; for d in 0..4 { let adj = ADJACENTS[d]; let next = p + adj; if next == q { dir = d; break; } } if !state.map[p][dir] { return Action::Construct(p, q); } } Action::Money } const MAX_HIGHWAY: usize = 65; #[derive(Debug, Clone)] struct AnnealingState { map: Map2d<[bool; 4]>, count: usize, candidates: Vec<(Coordinate, Coordinate)>, } impl AnnealingState { fn new() -> Self { let mut map = Map2d::new(vec![[false; 4]; N * N], N); for &row in &[3, 6, 10] { for col in 3..10 { let c = Coordinate::new(row, col); map[c][1] = true; map[c + ADJACENTS[1]][3] = true; } } for &col in &[3, 6, 10] { for row in 3..10 { let c = Coordinate::new(row, col); map[c][2] = true; map[c + ADJACENTS[2]][0] = true; } } let mut count = 0; for row in 0..N { for col in 0..N { for &dir in &[1, 2] { if map[Coordinate::new(row, col)][dir] { count += 1; } } } } Self { map, count, candidates: vec![], } } fn calc_score(&self, input: &Input) -> i64 { let mut distances = Map2d::new(vec![Map2d::new(vec![], N); N * N], N); let mut queue = BinaryHeap::new(); for &start in input.sampled_houses.iter() { let mut dists = Map2d::new(vec![std::i32::MAX / 2; N * N], N); queue.clear(); dists[start] = 0; queue.push(Reverse((0, start))); while let Some(Reverse((dist, c))) = queue.pop() { if dist > dists[c] { continue; } for dir in 0..4 { let next = c + ADJACENTS[dir]; let next_cost = dist + if self.map[c][dir] { 223 } else { 1000 }; if next.in_map(N) && dists[next].change_min(next_cost) { queue.push(Reverse((next_cost, next))); } } } distances[start] = dists; } let mut score = 0; for &person in &input.sampled_people { let dist = distances[person.home][person.company]; score += 60 * input.used_count_dict[dist as usize]; } score } fn can_flip(&self, target: Coordinate, dir: usize) -> bool { let next = target + ADJACENTS[dir]; next.in_map(N) && (self.count < MAX_HIGHWAY || self.map[target][dir]) } fn flip(&mut self, target: Coordinate, dir: usize) { let next = target + ADJACENTS[dir]; if self.map[target][dir] { self.count -= 1; } else { self.count += 1; } self.map[target][dir] ^= true; self.map[next][inv(dir)] ^= true; } fn update_order(&mut self) { let mut candidates = vec![]; for row in 0..N { for col in 0..N { let c = Coordinate::new(row, col); for &dir in &[1, 2] { if self.map[c][dir] { candidates.push((c, c + ADJACENTS[dir])); } } } } let mut dists = Map2d::new(vec![std::i32::MAX / 2; N * N], N); let mut queue = BinaryHeap::new(); let start = Coordinate::new(6, 6); dists[start] = 0; queue.push(Reverse((0, start))); while let Some(Reverse((dist, c))) = queue.pop() { if dist > dists[c] { continue; } for dir in 0..4 { let next = c + ADJACENTS[dir]; let next_cost = dist + if self.map[c][dir] { 1 } else { 5 }; if next.in_map(N) && dists[next].change_min(next_cost) { queue.push(Reverse((next_cost, next))); } } } candidates.sort_unstable_by_key(|&(p, q)| dists[p].min(dists[q])); self.candidates = candidates; } } trait Neighbor { fn apply(&self, input: &Input, state: &mut AnnealingState); fn rollback(&self, input: &Input, state: &mut AnnealingState); fn can_apply(&self, input: &Input, state: &AnnealingState) -> bool; } struct FlipOne { target: Coordinate, dir: usize, } impl FlipOne { fn new(target: Coordinate, dir: usize) -> Self { Self { target, dir } } } impl Neighbor for FlipOne { fn apply(&self, input: &Input, state: &mut AnnealingState) { state.flip(self.target, self.dir); } fn rollback(&self, input: &Input, state: &mut AnnealingState) { state.flip(self.target, self.dir); } fn can_apply(&self, input: &Input, state: &AnnealingState) -> bool { state.can_flip(self.target, self.dir) } } struct FlipTwo { target0: Coordinate, dir0: usize, target1: Coordinate, dir1: usize, } impl FlipTwo { fn new(target0: Coordinate, dir0: usize, target1: Coordinate, dir1: usize) -> Self { Self { target0, dir0, target1, dir1, } } } impl Neighbor for FlipTwo { fn apply(&self, input: &Input, state: &mut AnnealingState) { state.flip(self.target0, self.dir0); state.flip(self.target1, self.dir1); } fn rollback(&self, input: &Input, state: &mut AnnealingState) { state.flip(self.target0, self.dir0); state.flip(self.target1, self.dir1); } fn can_apply(&self, input: &Input, state: &AnnealingState) -> bool { !(self.target0 == self.target1 && self.dir0 == self.dir1) && (state.map[self.target0][self.dir0] ^ state.map[self.target1][self.dir1]) } } fn gen_target(rng: &mut Xoshiro256) -> (Coordinate, usize) { loop { let target = Coordinate::new(rng.gen_usize(2, N - 2), rng.gen_usize(2, N - 2)); let dir = rng.gen_usize(1, 3); if !(target.row == N - 1 && dir == 2) && !(target.col == N - 1 && dir == 1) { return (target, dir); } } } fn gen_neighbor(rng: &mut Xoshiro256) -> Box { if rng.gen_bool(0.2) { let (target, dir) = gen_target(rng); Box::new(FlipOne::new(target, dir)) } else { let (target0, dir0) = gen_target(rng); let (target1, dir1) = gen_target(rng); Box::new(FlipTwo::new(target0, dir0, target1, dir1)) } } fn annealing(input: &Input, initial_solution: AnnealingState, duration: f64) -> AnnealingState { let mut solution = initial_solution; let mut best_solution = solution.clone(); let mut current_score = solution.calc_score(input); let mut best_score = current_score; let init_score = current_score; let mut all_iter = 0; let mut valid_iter = 0; let mut accepted_count = 0; let mut update_count = 0; let mut rng = Xoshiro256::new(42); let duration_inv = 1.0 / duration; let since = std::time::Instant::now(); let temp0 = 1e4; let temp1 = 1e3; let mut inv_temp = 1.0 / temp0; loop { all_iter += 1; if (all_iter & ((1 << 4) - 1)) == 0 { let time = (std::time::Instant::now() - since).as_secs_f64() * duration_inv; let temp = f64::powf(temp0, 1.0 - time) * f64::powf(temp1, time); inv_temp = 1.0 / temp; if time >= 1.0 { break; } } // 変形 let neighbor = gen_neighbor(&mut rng); if !neighbor.can_apply(input, &solution) { continue; } neighbor.apply(input, &mut solution); // スコア計算 let new_score = solution.calc_score(input); let score_diff = new_score - current_score; if score_diff >= 0 || rng.gen_bool(f64::exp(score_diff as f64 * inv_temp)) { // 解の更新 current_score = new_score; accepted_count += 1; if best_score.change_max(current_score) { best_solution = solution.clone(); update_count += 1; } } else { neighbor.rollback(input, &mut solution); } valid_iter += 1; } eprintln!("===== annealing ====="); eprintln!("init score : {}", init_score); eprintln!("score : {}", best_score); eprintln!("all iter : {}", all_iter); eprintln!("valid iter : {}", valid_iter); eprintln!("accepted : {}", accepted_count); eprintln!("updated : {}", update_count); eprintln!(""); best_solution } mod judge { use std::{ cmp::Reverse, collections::{BinaryHeap, HashSet}, io::{stdout, Write}, }; use crate::{ grid::{inv, Coordinate, Map2d, ADJACENTS}, Action, ChangeMinMax, Input, State, N, }; pub(crate) trait Judge { fn read_input(&mut self) -> Input; fn update_state(&mut self, state: &mut State); fn apply(&mut self, action: Action); } pub(crate) struct OnlineJudge; impl Judge for OnlineJudge { fn read_input(&mut self) -> Input { Input::read_input() } fn update_state(&mut self, state: &mut State) { let (money, collaborator) = get!(i64, i64); if money == -1 && collaborator == -1 { panic!(); } state.money = money; state.collaborator = collaborator; } fn apply(&mut self, action: Action) { println!("{}", action); stdout().flush().unwrap(); } } pub(crate) struct SelfJudge { input: Input, state: State, turn: usize, } impl SelfJudge { pub(crate) fn new() -> Self { let input = Input { n: 0, t: 0, people: vec![], sampled_people: vec![], sampled_houses: HashSet::new(), used_count_dict: vec![], }; let state = State::init(); Self { input, state, turn: 0, } } } impl Judge for SelfJudge { fn read_input(&mut self) -> Input { self.input = Input::read_input(); self.input.clone() } fn update_state(&mut self, state: &mut State) { self.turn += 1; eprintln!("[TURN {:3}]", self.turn); let mut distances = Map2d::new(vec![Map2d::new(vec![], N); N * N], N); let mut queue = BinaryHeap::new(); for row in 0..N { for col in 0..N { let start = Coordinate::new(row, col); let mut dists = Map2d::new(vec![std::i32::MAX / 2; N * N], N); queue.clear(); dists[start] = 0; queue.push(Reverse((0, start))); while let Some(Reverse((dist, c))) = queue.pop() { if dist > dists[c] { continue; } for dir in 0..4 { let next = c + ADJACENTS[dir]; let next_cost = dist + if self.state.map[c][dir] { 223 } else { 1000 }; if next.in_map(N) && dists[next].change_min(next_cost) { queue.push(Reverse((next_cost, next))); } } } distances[start] = dists; } } for &person in &self.input.people { let dist = distances[person.home][person.company]; self.state.money += 60 * self.input.used_count_dict[dist as usize]; } eprintln!("money: {}", self.state.money); state.money = self.state.money; state.collaborator = self.state.collaborator; } fn apply(&mut self, action: Action) { eprintln!("action: {}", action); match action { Action::Construct(p, q) => { let cost = self.state.calc_construction_cost(); if self.state.money < cost { panic!(); } self.state.money -= cost; for dir in 0..4 { if p + ADJACENTS[dir] == q { self.state.map[p][dir] = true; self.state.map[q][inv(dir)] = true; return; } } } Action::Collaboration => { self.state.collaborator += 1; } Action::Money => { self.state.money += 50000; } } } } } #[allow(dead_code)] pub mod grid { use std::ops::{Div, DivAssign, Mul, MulAssign}; #[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)] pub struct Coordinate { pub row: usize, pub col: usize, } impl Coordinate { pub const fn new(row: usize, col: usize) -> Self { Self { row, col } } pub fn in_map(&self, size: usize) -> bool { self.row < size && self.col < size } pub const fn to_index(&self, size: usize) -> usize { self.row * size + self.col } pub const fn dist(&self, other: &Self) -> usize { Self::dist_1d(self.row, other.row) + Self::dist_1d(self.col, other.col) } const fn dist_1d(x0: usize, x1: usize) -> usize { (x0 as i64 - x1 as i64).abs() as usize } } impl Mul for Coordinate { type Output = Coordinate; fn mul(self, rhs: usize) -> Self::Output { Self::new(self.row * rhs, self.col * rhs) } } impl MulAssign for Coordinate { fn mul_assign(&mut self, rhs: usize) { self.row *= rhs; self.col *= rhs; } } impl Div for Coordinate { type Output = Coordinate; fn div(self, rhs: usize) -> Self::Output { Self::new(self.row / rhs, self.col / rhs) } } impl DivAssign for Coordinate { fn div_assign(&mut self, rhs: usize) { self.row /= rhs; self.col /= rhs; } } #[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)] pub struct CoordinateDiff { pub dr: usize, pub dc: usize, } impl CoordinateDiff { pub const fn new(dr: usize, dc: usize) -> Self { Self { dr, dc } } pub const fn invert(&self) -> Self { Self::new(0usize.wrapping_sub(self.dr), 0usize.wrapping_sub(self.dc)) } } impl std::ops::Add for Coordinate { type Output = Coordinate; fn add(self, rhs: CoordinateDiff) -> Self::Output { Coordinate::new(self.row.wrapping_add(rhs.dr), self.col.wrapping_add(rhs.dc)) } } pub const ADJACENTS: [CoordinateDiff; 4] = [ CoordinateDiff::new(!0, 0), CoordinateDiff::new(0, 1), CoordinateDiff::new(1, 0), CoordinateDiff::new(0, !0), ]; pub const DIRECTIONS: [char; 4] = ['U', 'R', 'D', 'L']; pub fn inv(dir: usize) -> usize { dir ^ 2 } #[derive(Debug, Clone)] pub struct Map2d { pub width: usize, pub height: usize, map: Vec, } impl Map2d { pub fn new(map: Vec, width: usize) -> Self { let height = map.len() / width; debug_assert!(width * height == map.len()); Self { width, height, map } } } impl std::ops::Index for Map2d { type Output = T; #[inline] fn index(&self, coordinate: Coordinate) -> &Self::Output { &self.map[coordinate.row * self.width + coordinate.col] } } impl std::ops::IndexMut for Map2d { #[inline] fn index_mut(&mut self, coordinate: Coordinate) -> &mut Self::Output { &mut self.map[coordinate.row * self.width + coordinate.col] } } impl std::ops::Index<&Coordinate> for Map2d { type Output = T; #[inline] fn index(&self, coordinate: &Coordinate) -> &Self::Output { &self.map[coordinate.row * self.width + coordinate.col] } } impl std::ops::IndexMut<&Coordinate> for Map2d { #[inline] fn index_mut(&mut self, coordinate: &Coordinate) -> &mut Self::Output { &mut self.map[coordinate.row * self.width + coordinate.col] } } impl std::ops::Index for Map2d { type Output = [T]; #[inline] fn index(&self, row: usize) -> &Self::Output { let begin = row * self.width; let end = begin + self.width; &self.map[begin..end] } } impl std::ops::IndexMut for Map2d { #[inline] fn index_mut(&mut self, row: usize) -> &mut Self::Output { let begin = row * self.width; let end = begin + self.width; &mut self.map[begin..end] } } } mod rand { pub(crate) struct Xoshiro256 { s0: u64, s1: u64, s2: u64, s3: u64, } impl Xoshiro256 { pub(crate) fn new(mut seed: u64) -> Self { let s0 = split_mix_64(&mut seed); let s1 = split_mix_64(&mut seed); let s2 = split_mix_64(&mut seed); let s3 = split_mix_64(&mut seed); Self { s0, s1, s2, s3 } } fn next(&mut self) -> u64 { let result = (self.s1 * 5).rotate_left(7) * 9; let t = self.s1 << 17; self.s2 ^= self.s0; self.s3 ^= self.s1; self.s1 ^= self.s2; self.s0 ^= self.s3; self.s2 ^= t; self.s3 = self.s3.rotate_left(45); result } pub(crate) fn gen_usize(&mut self, lower: usize, upper: usize) -> usize { assert!(lower < upper); let count = upper - lower; (self.next() % count as u64) as usize + lower } pub(crate) fn gen_i32(&mut self, lower: i32, upper: i32) -> i32 { assert!(lower < upper); let count = upper - lower; (self.next() % count as u64) as i32 + lower } pub(crate) fn gen_f64(&mut self) -> f64 { const UPPER_MASK: u64 = 0x3ff0000000000000; const LOWER_MASK: u64 = 0xfffffffffffff; let result = UPPER_MASK | (self.next() & LOWER_MASK); let result: f64 = unsafe { std::mem::transmute(result) }; result - 1.0 } pub(crate) fn gen_bool(&mut self, prob: f64) -> bool { self.gen_f64() < prob } } fn split_mix_64(x: &mut u64) -> u64 { *x += 0x9e3779b97f4a7c15; let mut z = *x; z = (z ^ z >> 30) * 0xbf58476d1ce4e5b9; z = (z ^ z >> 27) * 0x94d049bb133111eb; return z ^ z >> 31; } }