ICode9

精准搜索请尝试: 精确搜索
首页 > 其他分享> 文章详细

倍增,DFS序,欧拉序和树的一些知识

2022-08-10 20:32:29  阅读:61  来源: 互联网

标签:pre dist int rep cin DFS 序和树 void 欧拉


倍增

定义

倍增法,顾名思义就是翻倍.

它能够使线性的处理转化为对数级的处理,大大地优化时间复杂度

这个方法在很多算法中均有应用,其中最常用的是 RMQ 问题和求LCA,无修改的路径信息。

路径最小值

注意:路径上的信息需要可以合并,例如求最值

const int N = 201000;
const int LOGN = 18;

int n, q;
int dep[N], par[N][LOGN + 1], val[N][LOGN + 1];
vector< pair<int,int> > e[N];

il void dfs(int u, int f)
{
  dep[u] = dep[f] + 1;
  for (auto p : e[u])
  {
    int v = p.fi;
    if (v == f) continue;
    par[v][0] = u;
    val[v][0] = p.second;
    dfs(v, u);
  }
}

int query(int u, int v)
{
  int ans = 1 << 30;
  if (dep[u] > dep[v]) swap(u, v);
  int d = dep[v] - dep[u];
  per(j,LOGN,0) if (d & (1 << j))
  {
    ans = min(ans, val[v][j]);
    v = par[v][j];
  }
  if (u == v) return ans;
  per(j,LOGN,0) if (par[u][j] != par[v][j])
  {
    ans = min(ans, min(val[u][j], val[v][j]));
    u = par[u][j];
    v = par[v][j];
  }
  ans = min({ans, val[u][0], val[v][0]});
  return ans;
}


int main(void) 
{
  clock_t c1 = clock();
  FastIO();
  cin >> n >> q;
  rep(i,1,n-1)
  {
    int u, v, w;
    cin >> u >> v >> w;
    e[u].pb(make_pair(v,w));
    e[v].pb(make_pair(u,w));
  } 
  dfs(1,0);
  //预处理掉这部分数据
  rep(j,1,LOGN) rep(u,1,n) 
  {
    par[u][j] = par[par[u][j-1]][j-1];
    val[u][j] = min(val[u][j-1], val[par[u][j-1]][j-1]);
  }
  rep(i,1,q)
  {
    int u, v;
    cin >> u >> v;
    cout << query(u, v) << '\n';
  }
  cerr << "Time Used:" << clock() - c1 << "ms" << endl;
  return 0;
}

DFS序

定义

树的DFS序列,也就是树的深搜序,它的概念是:树的每一个节点在深度优先遍历中进出栈的时间序列。

作用

可以把子树问题转化为序列问题,非线性问题变成线性问题求解,可以用上树状数组或者线段树等数据结构进行信息维护。

可以用来做带修改的子树信息这一类题目。

例题1

题面

给一颗n个点的树,有两个操作。

  1. 1 x y, 把x的点权修改成y
  2. 2 x,询问x点的子树点权和和到根的路径的点权和(含x点)

代码实现

两个树状数组,分别维护前缀和和差分。

前缀和数组用来维护子树点权和,差分数组用来维护到根的路径点全和。

const int N = 2e5 + 10;

template<class T>
struct BIT
{
  T c[N];
  int size;
  void resize(int s) 
  {
    size = s;
    rep(i,1,size) c[i] = 0;
  }
  void modify(int x, T d)
  {
    assert(x != 0);
    for (; x <= size; x += x & (-x)) c[x] += d;
  }
  T query(int x)
  {
    assert(x<=size);
    T s = 0;
    for (; x; x -= x & (-x)) s += c[x];
    return s;
  }
};

int n, q, tot;
int a[N],l[N],r[N];
vector<int> e[N];
BIT<ll> c1, c2;

il void dfs(int u, int f) 
{
  l[u] = ++tot;
  for (auto v : e[u]) if (v == f) continue; else dfs(v, u);
  r[u] = tot;
}

int main(void) 
{
  // clock_t c1 = clock();
  FastIO();

  cin >> n >> q;
  rep(i,2,n) 
  {
    int u, v;
    cin >> u >> v;
    e[u].pb(v), e[v].pb(u);
  }

  dfs(1, 0);
  c1.resize(n), c2.resize(n);
  rep(i,1,n) 
  {
    cin >> a[i];
    c1.modify(l[i], a[i]);
    c2.modify(l[i], a[i]);
    c2.modify(r[i] + 1, -a[i]);
  }
  
  rep(i,1,q)
  {
    int ty; cin >> ty;
    if (ty == 1)
    {
      int x, y;
      cin >> x >> y;
      int d = y - a[x];
      a[x] = y;
      c1.modify(l[x], d);
      c2.modify(l[x], d);
      c2.modify(r[x] + 1, -d);
    }
    else
    {
      int x;
      cin >> x;
      cout << c1.query(r[x]) - c1.query(l[x] - 1) << " ";
      cout << c2.query(l[x]) << '\n';
    }
  }
  // cerr << "Time Used:" << clock() - c1 << "ms" << endl;
  return 0;
}

DFS序列2

代码实现

const int N = 2e5 + 10;

template<class T>
struct BIT
{
  T c[N];
  int size;
  void resize(int s) 
  {
    size = s;
    rep(i,1,size) c[i] = 0;
  }
  void modify(int x, T d)
  {
    assert(x != 0);
    for (; x <= size; x += x & (-x)) c[x] += d;
  }
  T query(int x)
  {
    assert(x<=size);
    T s = 0;
    for (; x; x -= x & (-x)) s += c[x];
    return s;
  }
};

int n, q, tot;
int a[N],l[N],r[N];
vector<int> e[N];
vector< pair<int, int> > son[N];
BIT<ll> c1;

void dfs(int u, int f) 
{
  l[u] = ++tot;
  for (auto v : e[u]) 
  {
    if (v == f) continue; 
    else dfs(v, u);
    son[u].pb({l[v], r[v]});
  }
  r[u] = tot;
}

int main(void) 
{
  // clock_t c1 = clock();
  ios::sync_with_stdio(false), cin.tie(nullptr), cout.tie(nullptr);

  cin >> n >> q;
  rep(i,2,n) 
  {
    int u, v;
    cin >> u >> v;
    e[u].pb(v), e[v].pb(u);
  }
  int root = 1;
  dfs(1, 0);
  c1.resize(n);
  rep(i,1,n) 
  {
    cin >> a[i];
    c1.modify(l[i], a[i]);
  }
  
  rep(i,1,q)
  {
    int ty; cin >> ty;
    if (ty == 1)
    {
      int x, y;
      cin >> x >> y;
      int d = y - a[x];
      a[x] = y;
      c1.modify(l[x], d);
    }
    else if (ty == 3)
    {
      cin >> root;
    }
    else
    {
      int x;
      cin >> x;
      if (x == root) cout << c1.query(n) << '\n';
      else if (l[x] < l[root] && r[root] <= r[x])
      {
        auto seg = *prev(upper_bound(all(son[x]), pair<int,int>{l[root], r[root]}));
        cout << c1.query(n) - (c1.query(seg.second) - c1.query(seg.first - 1)) << '\n';
      }
      else cout << c1.query(r[x]) - c1.query(l[x] - 1) << '\n';
    }
  }
  return 0;
}

欧拉序

相比DFS序,多记录一层回溯的点。

const int N = 501000, LOGN = 20;

int n, q, p[N], tot, dep[N];
vector<int> e[N];
PII f[LOGN + 2][N * 2];
ll ans;

unsigned int A, B, C;
inline unsigned int rng61() {
    A ^= A << 16;
    A ^= A >> 5;
    A ^= A << 1;
    unsigned int t = A;
    A = B;
    B = C;
    C ^= t ^ A;
    return C;
}

void dfs(int u, int par) 
{
  p[u] = ++tot;
  dep[u] = dep[par] + 1;
  f[0][tot] = {dep[u], u}; 
  for (auto v : e[u])
  {
    if (v == par) continue;
    dfs(v, u);
    f[0][++tot] = {dep[u], u};
  }
}

int main(){
  scanf("%d%d%u%u%u", &n, &q, &A, &B, &C);
  for (int i = 1; i < n; i++) 
  {
    int u, v;
    scanf("%d%d",&u,&v);
    e[u].pb(v);
    e[v].pb(u);
  }

  dfs(1, 0);
	for (int j = 1; j <= LOGN; j++) 
		for (int i = 1; i + (1 << j) - 1 <= tot; i++) 
			f[j][i] = min(f[j-1][i], f[j - 1][i + (1 << (j - 1))]);

  for (int i = 1; i <= q; i++) 
  {
    int l = p[rng61() % n + 1], r = p[rng61() % n + 1];
    if (l > r) swap(l, r);
		int len = 31 - __builtin_clz(r - l + 1);
    int d = min(f[len][l], f[len][r - (1 << len) + 1]).second;
    ans ^= 1ll * i * d;
  }
	printf("%lld\n",ans);
	return 0;
}

树的直径

树的直径是指树上任意两个节点之间最长.

树的直径的中间节点被称为树的中心,如果直径上有偶数个点,那么中间的两个节点都可以是树的中心.

树的中心到其他点的最长路径最短.

代码实现

int n, pre[N], c[N], q[N], dist[N];
int l, front = 1, rear = 0;
vector<int> e[N];

il void dfs(int x)
{
  for (auto y : e[x])
  {
    if (y != pre[x])
    {
      pre[y] = x;
      dist[y] = dist[x] + 1;
      dfs(y);
    }
  }
}

int main(void) 
{
  FastIO();
  
  cin >> n;
  rep(i,1,n-1)
  {
    int u, v;
    cin >> u >> v;
    e[u].pb(v), e[v].pb(u);
  }
  memset(dist, 0, sizeof(dist));
  memset(pre, 0, sizeof(pre));
  pre[1] = -1;
  dfs(1);
  int idx = 0, v = 0;
  rep(i,1,n)
  {
    if (dist[i] > v) v = dist[i], idx = i;
  }
  memset(dist, 0, sizeof(dist));
  memset(pre, 0, sizeof(pre));
  pre[idx] = -1;
  dfs(idx);
  v = 0;
  rep(i,1,n) v = max(v, dist[i]);
  cout << v << '\n';
  return 0;
}

树的重心

对于一颗无根树而言,当一个节点被选为根节点,它底下的每一个子节点的子树的大小最大值最小的那个点,被称为树的重心

删除重心后,树分裂成若干个子树,这若干个子树中的最大值最小.

性质

  • 当重心为根节点时,它底下的每一个子树大小不大于整棵树大小的一半
  • 重心到其他所有节点的距离和最小

代码实现

小数据\(O(logn)\)

const int N = 1e5 + 10;

int n, pre[1001], c[N], q[N], dist[N], f[N];
int l, front = 1, rear = 0;
vector<int> e[N];

il void dfs(int x)
{
  for (auto y : e[x])
  {
    if (y != pre[x])
    {
      pre[y] = x;
      dist[y] = dist[x] + 1;
      dfs(y);
    }
  }
}

il void solve(int x)
{
  ++cnt;
  for (auto y : e[x]) 
    if (y != pre[x]) 
      pre[y] = x, solve(y);
}

int main(void) 
{
  FastIO();
  
  cin >> n;
  rep(i,1,n-1)
  {
    int u, v;
    cin >> u >> v;
    e[u].pb(v), e[v].pb(u);
  }
  rep(i,1,n)
  {
    f[i] = 0;
    memset(pre, 0, sizeof(pre));
    for (auto y : e[i]) 
    {
      cnt = 0;
      pre[y] = i;
      solve(y);
      f[i] = max(f[i], cnt);
    }
  }
  // idx 重心
  int idx = 0, v = 1 << 30;
  rep(i,1,n) if (f[i] < v) v = f[i], idx = i;

  memset(dist, 0, sizeof(dist));
  memset(pre, 0, sizeof(pre));
  pre[idx] = -1;
  dfs(idx);
  int ans = 0;
  rep(i,1,n) ans += dist[i];
  cout << ans << '\n';
  return 0;
}

大数据

时间复杂度\(O(logn)\)

#include <bits/stdc++.h>
using namespace std;
#define DABIAO freopen("in.in","r",stdin);freopen("out.out","w",stdout);
#define debug(x...) do { cout << "\033[32;1m" << #x << " --> "; rd_debug(x); } while (0)
void rd_debug() {cout << "\033[39;0m" << endl; }
template<class T, class... Ts> void rd_debug(const T& arg, const Ts&... args) { cout << arg << " "; rd_debug(args...); }
#define FastIO() ios::sync_with_stdio(false),cin.tie(nullptr),cout.tie(nullptr)
#define rep(i, l, r) for(int i = (l); i <= (r); i++)
#define per(i, r, l) for(int i = (r); i >= (l); i--)
#define pb push_back
#define SZ(x) ((int)(x).size())
#define fi first
#define se second
#define all(x) (x).begin(), (x).end()
#define il inline
typedef long long ll;
typedef double db;
typedef pair<int, int> PII;

const int inf = 0x3f3f3f3f;
const ll infl = 0x3f3f3f3f3f3f3f3fll;
const double eps = 1e-6;
const int N = 1e5 + 10;

int n, pre[1001], c[N], q[N], dist[N], f[N];
int l, front = 1, rear = 0;
vector<int> e[N];

il void dfs(int x)
{
  for (auto y : e[x])
  {
    if (y != pre[x])
    {
      pre[y] = x;
      dist[y] = dist[x] + 1;
      dfs(y);
    }
  }
}

il void solve(int x)
{
  s[x] = 1;
  for (auto y : e[x]) 
    if (y != pre[x]) 
      pre[y] = x, solve(y), s[x] += s[y];
}

int main(void) 
{
  FastIO();
  
  cin >> n;
  rep(i,1,n-1)
  {
    int u, v;
    cin >> u >> v;
    e[u].pb(v), e[v].pb(u);
  }
  memset(pre, 0, sizeof(pre));
  pre[1] = -1;
  solve(1);

  int idx = 0, v = 1 << 30;
  rep(i,1,n)
  {
    int f = 0;
    for (auto y : e[i])
    {
      if (y != pre[i])
        f = max(f, s[y]);
      else 
        f = max(f, n - s[i]);
      if (f < v) v = f, idx = i;
    }
  }

  memset(dist, 0, sizeof(dist));
  memset(pre, 0, sizeof(pre));
  pre[idx] = -1;
  dfs(idx);
  int ans = 0;
  rep(i,1,n) ans += dist[i];
  cout << ans << '\n';
  return 0;
}		

LCA

求法

  1. 先计算两个节点深度
  2. 调整同一深度
  3. 两个节点一起往上跳,直到两个节点相等

代码实现

小数据

时间复杂度\(O(n^2)\)

const int N = 1001;
int n, m, father[N], dist[N];
vector<int> e[N];

il void dfs(int x){  for (auto y : e[x]) dist[y] = dist[x] + 1, dfs(y);  }

int main(void) 
{
  clock_t c1 = clock();
  FastIO();
  cin >> n;
  rep(i,1,n) 
  {
    int x, y;
    cin >> x >> y;
    e[x].pb(y);
    father[y] = x;
  }
  memset(dist, 0, sizeof(dist));
  dfs(1);
  cin >> m;
  rep(i,1,m)
  {
    int x, y;
    cin >> x >> y;
    if (dist[x] < dist[y]) swap(x, y);
    int z = dist[x] - dist[y];
    rep(i,j,z) x = father[x];
    while (x != y) x = father[x], y = father[y];
    cout << x << '\n';
  }
  cerr << "Time Used:" << clock() - c1 << "ms" << endl;
  return 0;
}

大数据

时间复杂度\(O(logn)\)

int n, m, father[N], dist[N];
vector<int> e[N];

il void dfs(int x)
{
  for (auto y : e[x]) dist[y] = dist[x] + 1, dfs(y);
}

int main(void) 
{
  clock_t c1 = clock();
  FastIO();
  cin >> n;
  rep(i,1,n) 
  {
    int x, y;
    cin >> x >> y;
    e[x].pb(y);
    father[y][0] = x;
  }
  rep(i,1,20) rep(j,1,n)
    if (father[j][i - 1])
      father[j][i] = father[father[j][i-1]][i-1];
  memset(dist, 0, sizeof(dist));
  dfs(1);
  cin >> m;
  rep(i,1,m)
  {
    int x, y;
    cin >> x >> y;
    if (dist[x] < dist[y]) swap(x, y);
    int z = dist[x] - dist[y];
    for (int j = 0; j <= 20 && z; j++, z >>= 1)
      if (z & 1) x = father[x][j];
    if (x == y) {  cout << x << '\n'; continue;  }
    per(j,20,0) if (father[x][j] != father[y][j])
      x = father[x][j], y = father[y][j];
    cout << father[x][0] << '\n';
  }

  cerr << "Time Used:" << clock() - c1 << "ms" << endl;
  return 0;
}

用欧拉序来求LCA

const int N = 501000, LOGN = 20;

int n, q, p[N], tot, dep[N];
vector<int> e[N];
PII f[LOGN + 2][N * 2];
ll ans;

unsigned int A, B, C;
inline unsigned int rng61() {
    A ^= A << 16;
    A ^= A >> 5;
    A ^= A << 1;
    unsigned int t = A;
    A = B;
    B = C;
    C ^= t ^ A;
    return C;
}

void dfs(int u, int par) 
{
  p[u] = ++tot;
  dep[u] = dep[par] + 1;
  f[0][tot] = {dep[u], u}; 
  for (auto v : e[u])
  {
    if (v == par) continue;
    dfs(v, u);
    f[0][++tot] = {dep[u], u};
  }
}

int main(){
  scanf("%d%d%u%u%u", &n, &q, &A, &B, &C);
  for (int i = 1; i < n; i++) 
  {
    int u, v;
    scanf("%d%d",&u,&v);
    e[u].pb(v);
    e[v].pb(u);
  }

  dfs(1, 0);
	for (int j = 1; j <= LOGN; j++) 
		for (int i = 1; i + (1 << j) - 1 <= tot; i++) 
			f[j][i] = min(f[j-1][i], f[j - 1][i + (1 << (j - 1))]);

  for (int i = 1; i <= q; i++) 
  {
    int l = p[rng61() % n + 1], r = p[rng61() % n + 1];
    if (l > r) swap(l, r);
		int len = 31 - __builtin_clz(r - l + 1);
    int d = min(f[len][l], f[len][r - (1 << len) + 1]).second;
    ans ^= 1ll * i * d;
  }
	printf("%lld\n",ans);
	return 0;
}

标签:pre,dist,int,rep,cin,DFS,序和树,void,欧拉
来源: https://www.cnblogs.com/guyuLihua/p/16573776.html

本站声明: 1. iCode9 技术分享网(下文简称本站)提供的所有内容,仅供技术学习、探讨和分享;
2. 关于本站的所有留言、评论、转载及引用,纯属内容发起人的个人观点,与本站观点和立场无关;
3. 关于本站的所有言论和文字,纯属内容发起人的个人观点,与本站观点和立场无关;
4. 本站文章均是网友提供,不完全保证技术分享内容的完整性、准确性、时效性、风险性和版权归属;如您发现该文章侵犯了您的权益,可联系我们第一时间进行删除;
5. 本站为非盈利性的个人网站,所有内容不会用来进行牟利,也不会利用任何形式的广告来间接获益,纯粹是为了广大技术爱好者提供技术内容和技术思想的分享性交流网站。

专注分享技术,共同学习,共同进步。侵权联系[81616952@qq.com]

Copyright (C)ICode9.com, All Rights Reserved.

ICode9版权所有