## 1. NFA的确定化
### 1.1. 明确NFA的定义
一个非确定的有穷自动机(NFA)M是一个五元式:
- M=(S,∑,δ,S0,F)
- S是一个有限集,它额每个元素称为一个状态。
- ∑是一个有穷字母表,它的每个元素称为一个输入字符
- δ是一个从S×∑∗至S子集额单值映射。即:δ:S×∑∗→2⋅S
- S0⊆S,是一个非空的初态集
- F⊂ S , 是一个终态集(可空)
### 1.2. 定义运算
定义对状态集合I的几个有关运算:
- 状态集合I的ε-闭包,表示为ε-closure(I),定义为一状态集,是状态集I中的任何状态s经任意条ε弧而能到达的状态的集合。状态集合I的任何状态s都属于ε-closure(I)。
- 状态集合I的a弧转换,表示为move(I,a)定义为状态集合J,其中J是所有那些可从I的某一状态经过一条a弧而到达的状态的全体。
定义Ia = ε-closure(move(I,a))
### 1.3. 算法描述
- 每次从队头取出一个集合,**(开始队列内只有初态集合I的ε-闭包(I) )**,然后得到它对于任意一个字符a的Ia=ε−closure(move(I,a))
- 然后如果当前状态之前没有出现过,那么当前状态作为一个新的状态I,放入队列。
- 一直做如上操作,直到队列为空
### 1.4. 代码如下:
~~~
#include <iostream>
#include <cstdio>
#include <algorithm>
#include <fstream>
#include <queue>
#include <vector>
#include <string>
#include <cstring>
#include <set>
#define MAX 500
#define M 1000007
using namespace std;
struct Set
{
set<int> elements;
int state;
}I[MAX];
vector<int> e[MAX];
vector<int> edge[MAX];
vector<int> val1[MAX];
vector<int> val2[MAX];
vector<int> hash[M];
vector<int> types;
int vis[MAX][MAX];
int cntp,cnts,start,finish,used[MAX];
void _clear( );
void clear( );
void _add( int u , int v , int x );
void add( int u , int v , int x );
void dfs ( set<int>& elements , int u , int d , int flag );
void ensure();
void minimize();
int get_hash( set<int>& item );
//#define DEBUG
int main ( )
{
int m;
puts("给出NFA的边的条数:");
while ( ~scanf ( "%d" , &m ) )
{
_clear();
clear();
puts("给出NFA的始态和终态:");
scanf ( "%d%d" , &start , &finish );
puts("给出所有的边,每行一条:");
for ( int i = 0 ; i < m ; i++ )
{
int u,v,x;
scanf ( "%d%d%d" , &u , &v , &x );
_add ( u , v , x );
}
#ifdef DEBUG
set<int> temp;
int num[]={2,3,6,7,8};
//for ( int i = 2 ; i < 6 ; i++ )
// dfs ( temp , i , 0 , 2 );
for ( int i = 0 ; i < 5 ; i++ )
dfs ( temp , num[i] , 0 , 1 );
set<int>::iterator it = temp.begin();
for ( ; it != temp.end() ; it++ )
printf ( "%d " , *it);
puts("");
#else
ensure();
puts("计算结果如下:");
printf ( "%d\n" , cnts );
for ( int i = 0 ; i < cnts ; i++ )
for ( int j = 0 ; j < edge[i].size() ; j++ )
printf ( "edges : %d %d %d\n" , i , edge[i][j] , val2[i][j] );
puts("\n给出DFA的边的条数:");
#endif
}
}
void _clear()
{
for ( int i = 0 ; i < MAX ; i++ )
{
e[i].clear();
val1[i].clear();
}
types.clear();
memset ( used , 0 , sizeof ( used ) );
}
void clear()
{
for ( int i = 0 ; i < MAX ; i++ )
{
edge[i].clear();
val2[i].clear();
}
}
void _add ( int u , int v , int x )
{
e[u].push_back ( v );
val1[u].push_back ( x );
if ( !used[x] )
{
types.push_back ( x );
used[x] = types.size();
}
}
int get_hash ( set<int>& item )
{
int sum = 0;
set<int>::iterator it = item.begin(),it1,it2;
for ( ; it!= item.end() ; it++ )
{
sum += (*it)*(*it)%M;
sum %= M;
}
for ( int i = 0 ; i < hash[sum].size() ; i++ )
{
int x = hash[sum][i];
set<int>& temp = I[x].elements;
it = temp.begin();
bool flag = true;
if ( temp.size() != item.size() ) continue;
for ( it1 = temp.begin() , it2 = item.begin(); it2!= item.end() ; it1++,it2++ )
if ( *it1 != *it2 )
{
flag = false;
break;
}
if ( flag ) return x;
}
return -1;
}
int make_hash ( set<int>& item )
{
int sum = 0;
set<int>::iterator it = item.begin();
for ( ; it!= item.end() ; it++ )
{
sum += (*it)*(*it)%M;
sum %= M;
}
hash[sum].push_back ( cnts );
return cnts++;
}
void add ( int u , int v , int x )
{
edge[u].push_back ( v );
val2[u].push_back ( x );
}
void dfs ( set<int>& elements , int u , int d , int flag )
{
if ( vis[u][d] ) return;
vis[u][d] = 1;
if ( d == 1 ) elements.insert ( u );
if ( d == 2 ) return;
int len = e[u].size();
for ( int i = 0 ; i < len ; i++ )
{
int v = e[u][i],dd;
int x = val1[u][i];
if ( x == flag ) dd = d+1;
else if ( x == 0 ) dd = d;
else continue;
dfs ( elements , v , dd , flag );
}
}
void ensure ( )
{
I[0].elements.insert(start);
cnts = 1;
for ( int i = 0 ; i < cnts ; i++ )
{
set<int>& cur = I[i].elements;
for ( int j = 0 ; j < types.size() ; j++ )
{
if ( types[j] == 0 ) continue;
memset ( vis , 0 , sizeof ( vis ) );
set<int>& temp = I[cnts].elements;
temp.clear();
set<int>::iterator it = cur.begin();
for ( ; it != cur.end() ; it++ )
dfs ( temp , *it , 0 , types[j] );
if ( temp.empty() ) continue;
int x = get_hash ( temp );
if ( x == -1 )
x = make_hash ( temp );
add ( i , x , types[j] );
}
set<int>::iterator it = cur.begin();
printf ( "I%d :" , i );
for ( ; it!= cur.end() ; it++ )
printf ( "%d " , *it );
puts ("");
}
}
~~~
## 2. DFA的最小化
### 2.1. 明确DFA的定义
一个确定的有穷自动机(DFA)M是一个五元式:
- M=(S, ∑, δ, s0, F)其中
- S是一个有限集,它的每个元素称为一个状态。
- ∑是一个有穷字母表,它的每个元素称为一个输入字符
-δ是一个从S×∑至S的单值映射。δ(s,a)=s’意味着:当现行状态-为s、输入字符为a时,将转换到下一个状态s’。我们称s’为s的一个后继状态。
- s0∈S,是唯一的初态。
- F ⊂ S,是一个终态集(可空)
### 2.2 算法描述
- DFA M =(K,∑,f, k0,, kt),最小状态DFA M’
- 1.构造状态的初始划分∏0:终态kt 和非终态K- kt两组
- 2.对∏施用传播性原则 构造新划分∏new
- 3.如∏new=∏,则令∏new=∏并继续步骤4,否则∏:=∏new重复2
- 4.为∏final中的每一组选一代表,这些代表构成M’的状态。若k是一代表且f(k,a)=t,令r是t组的代表,则M’中有一转换f’(k,a)=r M’ 的开始状态是含有K0的那组的代表 M’的终态是含有Kt的那组的代表
- 5.去掉M’中的死状态.
### 2.3 代码实现
~~~
#include <iostream>
#include <algorithm>
#include <cstring>
#include <cstdio>
#include <cstdlib>
#include <string>
#include <vector>
#include <map>
#include <queue>
#include <fstream>
#define MAX 507
using namespace std;
struct Node
{
int x;
int type;
Node(){}
Node( int a , int b )
:x(a),type(b){}
bool operator < ( const Node& a ) const
{
return type < a.type;
}
};
int mp[MAX][MAX];
vector<int> _edge[MAX];
vector<int> _val[MAX];
vector<int> edge[MAX];
vector<int> val[MAX];
vector<int> point;
vector<pair<int,int> > ans1;
vector<int> ans2;
vector<Node> col;
int fa[MAX];
int type[MAX];
int state[MAX];
int used[MAX];
int kinds;
int groups_num;
int n;
int m;
void init ();
int _find( int x );
void _union ( int x , int y );
void _add ( int u , int v , int x );
void add ( int u , int v , int x );
void minimize ( );
void make_csv();
int main ( )
{
init ( );
scanf ( "%d%d" , &n , &m );
for ( int i = 1 ; i <= n ; i++ )
scanf ( "%d" , &type[i] );
for ( int i = 0 ; i < m ; i++ )
{
int u,v,x;
scanf ( "%d%d%d" , &u , &v , &x );
_add ( u ,v , x );
}
minimize();
//最小化后的点的保留结果
puts ("points : " );
printf ( "%d\n" , (int)point.size() );
for ( int i = 0 ; i < point.size(); i++ )
printf ( "%d " , point[i] );
puts("");
//最小化后的边的保留结果
puts ( "edge: ");
m = ans1.size();
for ( int i = 0 ; i < ans1.size() ; i++ )
printf ( "%d %d %d\n" , ans1[i].first , ans1[i].second , ans2[i] );
puts("");
make_csv();
}
void init ( )
{
for ( int i = 0 ; i < MAX ; i++ )
{
edge[i].clear();
_edge[i].clear();
}
point.clear();
memset ( type , 255 , sizeof ( type ) );
memset ( used , 0 , sizeof ( used ) );
kinds = 0;
groups_num = 2;
col.clear();
for ( int i = 0 ; i < MAX ; i++ )
fa[i] = i;
}
void _add ( int u , int v , int x )
{
_edge[u].push_back ( v );
_val[u].push_back ( x );
if ( used[x] ) return ;
used[x] = x;
kinds++;
}
void add ( int u , int v , int x )
{
edge[u].push_back ( v );
val[u].push_back ( x );
}
int _find ( int x )
{
return x == fa[x]? x: fa[x] = _find ( fa[x] );
}
void _union ( int x , int y )
{
x = _find ( x );
y = _find ( y );
fa[x] = y;
}
void minimize ( )
{
queue<vector<Node> > q[2];
col.clear();
for ( int i = 1 ; i <= n ; i++ )
{
if ( type[i] == 0 )
col.push_back ( Node( i , 0 ) );
}
q[1].push ( col );
col.clear();
for ( int i = 1 ; i <= n ;i++ )
{
if ( type[i] == 1 )
col.push_back ( Node ( i , 1 ) );
}
q[1].push( col );
col.clear();
for ( int i = 1 ; i <= kinds ; i++ )
{
int cur = i%2;
int next = (i+1)%2;
while ( !q[next].empty() ) q[next].pop();
while ( !q[cur].empty() )
{
vector<Node> front = q[cur].front();
q[cur].pop();
for ( int j = 0 ; j < front.size() ; j++ )
{
Node& temp = front[j];
int u = temp.x;
for ( int k = 0 ; k < _edge[u].size() ; k++ )
{
int v = _edge[u][k];
int x = _val[u][k];
if ( x != i ) continue;
temp.type = type[v];
}
}
sort ( front.begin() , front.end() );
if ( front[0].type == front[front.size()-1].type )
q[next].push ( front );
else
{
col.clear();
col.push_back ( front[0] );
for ( int j = 1 ; j < front.size() ; j++ )
{
if ( front[j].type != front[j-1].type )
{
q[cur].push ( col );
col.clear();
groups_num++;
}
type[front[j].x] = groups_num;
col.push_back ( front[j] );
}
q[cur].push( col );
}
}
}
//#define DEBUG
#ifdef DEBUG
int id = (kinds+1)%2;
int num = 1;
while ( !q[id].empty() )
{
vector<Node> temp = q[id].front();
q[id].pop();
printf ( "%d : ", num++ );
for ( int i = 0 ; i < temp.size() ; i++ )
printf ( "%d " , temp[i].x );
puts("");
}
#else
int id = (kinds+1)%2;
while ( !q[id].empty() )
{
vector<Node> temp = q[id].front();
q[id].pop();
for ( int i = 1 ; i < temp.size() ; i++ )
_union ( temp[i].x , temp[i-1].x);
}
memset ( used , 0 , sizeof ( used ) );
memset ( mp , 0 , sizeof ( mp ) );
for ( int i = 1 ; i <= n ;i++ )
{
int x = _find(i);
if ( used[x] ) continue;
used[x] = 1;
point.push_back ( x );
}
ans1.clear();
for ( int i = 1 ; i <= n ; i++ )
for ( int j = 0 ; j < _edge[i].size() ; j++ )
{
int v = _edge[i][j];
int x = _val[i][j];
int a = _find(i);
int b = _find(v);
if ( mp[a][b] ) continue;
mp[a][b] = 1;
add ( a , b , x );
ans1.push_back ( make_pair ( a , b ) );
ans2.push_back ( x );
}
#endif
}
void make_csv ( )
{
freopen ( "g1.csv" , "w" , stdout );
printf ( "%d,%d, \n" , n , m );
for ( int i = 0 ; i < point.size() ; i++ )
if ( i == 0 ) printf ( "%d" , point[i] );
else printf ( ",%d" , point[i] );
puts("");
for ( int i = 0 ; i < m ; i++ )
printf ( "%d,%d,%d\n" , ans1[i].first , ans1[i].second , ans2[i] );
fclose(stdout);
}
~~~