You are given a connected weighted graph with n vertices and m edges. The graph doesn't contain loops nor multiple edges. Consider some edge with id i. Let's determine for this edge the maximum integer weight we can give to it so that it is contained in all minimum spanning trees of the graph if we don't change the other weights.
You are to determine this maximum weight described above for each edge. You should calculate the answer for each edge independently, it means there can't be two edges with changed weights at the same time.
The first line contains two integers n and m (2 ≤ n ≤ 2·105, n - 1 ≤ m ≤ 2·105), where n and m are the number of vertices and the number of edges in the graph, respectively.
Each of the next m lines contains three integers u, v and c (1 ≤ v, u ≤ n, v ≠ u, 1 ≤ c ≤ 109) meaning that there is an edge between vertices u and v with weight c.
Print the answer for each edge in the order the edges are given in the input. If an edge is contained in every minimum spanning tree with any weight, print -1 as the answer.
4 4
1 2 2
2 3 2
3 4 2
4 1 3
2 2 2 1
4 3
1 2 2
2 3 2
3 4 2
-1 -1 -1
题目大意 给定一个无向连通带权图,求每条边在所有最小生成树中的最大权值(如果可以无限大就输出-1)。
对于求1条边在无向连通图带权图的所有最小生成树中的最大权值可以用二分再用Kruskal进行check,但是如果每条边都这么做就会T掉。
所以考虑整体二分,然而并不行。
所以考虑先跑一遍Kruskal,然后分类讨论一下:
1)考虑一条非树边可以取的最大权值。
考虑把它加入树中,那么会形成1个环,为了保证最小生成树的边权和最小,方法是删掉环上权值最大的一条边。
所以找到它连接的两端在树上形成的简单路径中边权最大的一个,它的边权-1就是这条非树边的答案。
这个操作可以用树链剖分或者倍增解决。
2)考虑一条树边可以取到的最大权值
具体考虑一条树边会比较难做(不过好像有同学设计了时间戳把它搞定了),但是对于每条非树边都会对它连接的两端在树上形成的简单路径上的所有边有个边权的限制,就是不能超过它的边权 - 1,否则会被它替换掉。
这个区间取min操作可以用树链剖分。然而考试的时候我脑子瓦特了,觉得线段树不能区间取min(可能是脑补了一个求和操作)
然后想到了只有到最后会一起求得树边的答案,于是想到了差分。
为了维护这个最小值,又想到了可并堆。由于要删除,所以可以用下面这个方法构造可删堆(一位dalao的博客提到这个黑科技,我就学习了一下)
再开一个堆记录要删除的元素,如果两个堆堆顶元素相同,则都弹出堆顶元素。
于是便又有一个名为树差分 + 可并堆的zz做法。
Code
1 /** 2 * Codeforces 3 * Problem#828F 4 * Accepted 5 * Time: 420ms 6 * Memory: 57864k 7 */ 8 #include <iostream> 9 #include <fstream> 10 #include <sstream> 11 #include <cstdio> 12 #include <cstdlib> 13 #include <cstring> 14 #include <ctime> 15 #include <cmath> 16 #include <cctype> 17 #include <algorithm> 18 #include <map> 19 #include <set> 20 #include <queue> 21 #include <stack> 22 #include <vector> 23 #include <bitset> 24 #ifdef WIN32 25 #define Auto "%I64d" 26 #else 27 #define Auto "%lld" 28 #endif 29 using namespace std; 30 typedef bool boolean; 31 #define ll int 32 #define smin(_a, _b) _a = min(_a, _b) 33 #define smax(_a, _b) _a = max(_a, _b) 34 #define fi first 35 #define sc second 36 const signed int inf = (signed) (~0u >> 1); 37 const signed ll llf = (signed ll) (~0ull >> 1); 38 typedef pair<int, int> pii; 39 40 template<typename T> 41 inline void readInteger(T& u) { 42 static char x; 43 while(!isdigit(x = getchar())); 44 for(u = x - '0'; isdigit(x = getchar()); u = u * 10 + x - '0'); 45 } 46 47 template<typename T> 48 class Matrix { 49 public: 50 T* p; 51 int row; 52 int col; 53 Matrix():p(NULL) { } 54 Matrix(int row, int col):row(row), col(col) { 55 p = new T[(row * col)]; 56 } 57 58 T* operator [] (int pos) { 59 return p + (pos * col); 60 } 61 }; 62 #define matset(a, i, s) memset(a.p, i, sizeof(s) * a.row * a.col) 63 64 typedef class Edge { 65 public: 66 int u; 67 int v; 68 int w; 69 boolean seced; 70 int rid; 71 72 Edge(int u = 0, int v = 0, int w = 0):u(u), v(v), w(w), seced(false) { } 73 74 boolean operator < (Edge b) const { 75 return w < b.w; 76 } 77 }Edge; 78 79 typedef class Node { 80 public: 81 int val; 82 Node *nxt[2]; 83 84 Node(int val = inf):val(val) { nxt[0] = nxt[1] = NULL; } 85 }Node; 86 87 typedef pair<Node*, Node*> pnn; 88 #define limit 1000000 89 90 Node pool[limit]; 91 Node *top = pool; 92 93 Node* newnode(int x) { 94 if(top == pool + limit) 95 return new Node(x); 96 *top = Node(x); 97 return top++; 98 } 99 100 Node* merge(Node* &l, Node* r) { 101 if(l == NULL) return r; 102 if(r == NULL) return l; 103 if(l->val > r->val) swap(l, r); 104 int p = rand() % 2; 105 l->nxt[p] = merge(l->nxt[p], r); 106 return l; 107 } 108 109 int n, m; 110 Edge* edge; 111 int* res; 112 int* ans; 113 114 inline void init() { 115 readInteger(n); 116 readInteger(m); 117 edge = new Edge[(m + 1)]; 118 for(int i = 1; i <= m; i++) { 119 readInteger(edge[i].u); 120 readInteger(edge[i].v); 121 readInteger(edge[i].w); 122 edge[i].rid = i; 123 } 124 } 125 126 int *f; 127 128 int find(int x) { 129 return (f[x] != x) ? (f[x] = find(f[x])) : (x); 130 } 131 132 vector<int> *g; 133 vector<int> *add; 134 vector<int> *del; 135 136 inline void Kruskal() { 137 f = new int[(n + 1)]; 138 g = new vector<int>[(n + 1)]; 139 for(int i = 1; i <= n; i++) 140 f[i] = i; 141 sort(edge + 1, edge + m + 1); 142 int fin = 0; 143 for(int i = 1; i <= m && fin < n - 1; i++) { 144 if(find(edge[i].u) != find(edge[i].v)) { 145 f[find(edge[i].u)] = find(edge[i].v); 146 g[edge[i].u].push_back(i); 147 g[edge[i].v].push_back(i); 148 edge[i].seced = true; 149 fin++; 150 } 151 } 152 } 153 154 const int BZMAX = 18; 155 int* dep; 156 Matrix<int> bz; 157 Matrix<int> bzm; 158 159 void dfs1(int node, int fa, int lastv) { 160 dep[node] = dep[fa] + 1; 161 bz[node][0] = fa; 162 bzm[node][0] = lastv; 163 for(int i = 1; i < BZMAX; i++) 164 bz[node][i] = bz[bz[node][i - 1]][i - 1], bzm[node][i] = max(bzm[node][i - 1], bzm[bz[node][i - 1]][i - 1]); 165 for(int i = 0; i < (signed)g[node].size(); i++) { 166 if(!edge[g[node][i]].seced) continue; 167 int e = (edge[g[node][i]].u == node) ? (edge[g[node][i]].v) : (edge[g[node][i]].u); 168 if(e == fa) continue; 169 dfs1(e, node, edge[g[node][i]].w); 170 } 171 } 172 173 pii lca(int u, int v) { 174 if(dep[u] < dep[v]) swap(u, v); 175 int rtmax = 0; 176 int ca = dep[u] - dep[v]; 177 for(int i = 0; i < BZMAX; i++) 178 if(ca & (1 << i)) { 179 smax(rtmax, bzm[u][i]); 180 u = bz[u][i]; 181 } 182 if(u == v) return pii(u, rtmax); 183 for(int i = BZMAX - 1; ~i; i--) { 184 if(bz[u][i] != bz[v][i]) { 185 smax(rtmax, bzm[u][i]); 186 smax(rtmax, bzm[v][i]); 187 u = bz[u][i]; 188 v = bz[v][i]; 189 } 190 } 191 smax(rtmax, bzm[u][0]); 192 smax(rtmax, bzm[v][0]); 193 return pii(bz[u][0], rtmax); 194 } 195 196 pair<Node*, Node*> dfs2(int node, int fa, int tofa) { 197 Node* rt = NULL; 198 Node* dl = NULL; 199 for(int i = 0; i < (signed)g[node].size(); i++) { 200 if(!edge[g[node][i]].seced) continue; 201 int e = (edge[g[node][i]].u == node) ? (edge[g[node][i]].v) : (edge[g[node][i]].u); 202 if(e == fa) continue; 203 pnn ap = dfs2(e, node, g[node][i]); 204 rt = merge(rt, ap.fi); 205 dl = merge(dl, ap.sc); 206 } 207 for(int i = 0; i < (signed)add[node].size(); i++) 208 rt = merge(rt, newnode(add[node][i])); 209 for(int i = 0; i < (signed)del[node].size(); i++) 210 dl = merge(dl, newnode(del[node][i])); 211 while(dl && rt->val == dl->val) { 212 rt = merge(rt->nxt[0], rt->nxt[1]); 213 rt = merge(rt->nxt[0], rt->nxt[1]); 214 dl = merge(dl->nxt[0], dl->nxt[1]); 215 } 216 res[tofa] = (!rt) ? (-1) : (rt->val); 217 return pnn(rt, dl); 218 } 219 220 inline void solve() { 221 res = new int[(m + 1)]; 222 dep = new int[(n + 1)]; 223 bz = Matrix<int>(n + 1, BZMAX); 224 bzm = Matrix<int>(n + 1, BZMAX); 225 dep[0] = 0; 226 for(int i = 0; i < BZMAX; i++) 227 bz[0][i] = bzm[0][i] = 0; 228 dfs1(1, 0, 0); 229 add = new vector<int>[(n + 1)]; 230 del = new vector<int>[(n + 1)]; 231 for(int i = 1; i <= m; i++) { 232 if(edge[i].seced) continue; 233 int u = edge[i].u, v = edge[i].v; 234 pii l = lca(u, v); 235 res[i] = l.sc - 1; 236 add[u].push_back(edge[i].w - 1); 237 add[v].push_back(edge[i].w - 1); 238 del[l.fi].push_back(edge[i].w - 1); 239 } 240 dfs2(1, 0, 0); 241 ans = new int[(m + 1)]; 242 for(int i = 1; i <= m; i++) 243 ans[edge[i].rid] = res[i]; 244 for(int i = 1; i <= m; i++) 245 printf("%d ", ans[i]); 246 } 247 248 int main() { 249 srand(233); 250 init(); 251 Kruskal(); 252 solve(); 253 return 0; 254 }