[AC自动机+dp]PKU 3691--DNA repair

http://acm.pku.edu.cn/JudgeOnline/problem?id=3691

DNA repair

Time Limit: 2000MS		Memory Limit: 65536K
Total Submissions: 1469		Accepted: 650

Description

Biologists finally invent techniques of repairing DNA that contains segments causing kinds of inherited diseases. For the sake of simplicity, a DNA is represented as a string containing characters 'A', 'G' , 'C' and 'T'. The repairing techniques are simply to change some characters to eliminate all segments causing diseases. For example, we can repair a DNA "AAGCAG" to "AGGCAC" to eliminate the initial causing disease segments "AAG", "AGC" and "CAG" by changing two characters. Note that the repaired DNA can still contain only characters 'A', 'G', 'C' and 'T'.

You are to help the biologists to repair a DNA by changing least number of characters.

Input

The input consists of multiple test cases. Each test case starts with a line containing one integers N (1 ≤ N ≤ 50), which is the number of DNA segments causing inherited diseases.
The following N lines gives N non-empty strings of length not greater than 20 containing only characters in "AGCT", which are the DNA segments causing inherited disease.
The last line of the test case is a non-empty string of length not greater than 1000 containing only characters in "AGCT", which is the DNA to be repaired.

The last test case is followed by a line containing one zeros.

Output

For each test case, print a line containing the test case number( beginning with 1) followed by the
number of characters which need to be changed. If it's impossible to repair the given DNA, print -1.

 

Sample Input

2
AAA
AAG
AAAG    
2
A
TG
TGAATG
4
A
G
C
T
AGT
0

Sample Output

Case 1: 1
Case 2: 4
Case 3: -1

题目大意：给出一个字符串和若干个非法子串，求最少修改字符串中的多少个字符可以使得字符串中不存在非法子串。

分析：若字符串是01串，若子串长度为10，可以用经典的状态压缩dp解决，即：f[i,j]表示前i个，前9个数字状态为j的最小修改次数。然而，本题数据规模大得多，不能用这个dp。

我们来分析一下状态，对于以下的这个输入：

AAA

GT

XXXXXX

TC、AT、CT之类的状态其实都是一样的。

这些状态都是什么呢？

有穷状态自动机上的某些节点。

我们只要对所有的非法子串，构建一个AC自动机，那么自动机上的某个节点即为一类状态。

所以得到以下动态规划算法：

令f[i，j]表示对于字符串中的前i位，使得状态为自动机上的节点j，期间不经过任何一个非法子串需要修改的最少字符数。

则f[i,j]=min(f[i-1,k]+(1-ord(s[i]=way[k,j])))。其中，k为j的父亲节点（不一定是直接的父亲），way[k,j]为k连向j的边所代表的字母。这个转移方程不难理解，若第i位和边的字符不同，需要修改，答案+1，否则不用。答案便是max(f[len,i])，i不为某个非法子串的最后一个字母。

另外，在构建失败指针的同时，若某个节点的某个儿子为空，则要通过失败指针找到一个儿子，不然指向根。

具体的参看程序。

 

codes:

 

type tnode=record n:array[1..4] of longint; f:longint; d:boolean; end; const maxn=1000; var n,tot,ans,k:longint; b:array[1..maxn+1] of longint; t:array[0..maxn] of tnode; a:array[1..50] of string; f:array[0..1000,0..maxn] of longint; s:ansistring; function ch(a:char):longint; begin case a of 'A':ch:=1; 'G':ch:=2; 'C':ch:=3; 'T':ch:=4; end; end; procedure clr(x:longint); var i:longint; begin for i:=1 to 4 do t[x].n[i]:=0; t[x].d:=false; t[x].f:=0; end; procedure insert(x:string); var i,now,num:longint; begin now:=0; for i:=1 to length(x) do begin num:=ch(x[i]); if t[now].n[num]=0 then begin inc(tot); t[now].n[num]:=tot; clr(tot); end; now:=t[now].n[num]; end; t[now].d:=true; end; procedure build_ac; var now,temp,head,tail,i:longint; begin head:=0; tail:=1; b[1]:=0; now:=0; repeat inc(head); now:=b[head]; if not t[now].d then begin temp:=now; while (temp<>-1) and (not t[temp].d) do temp:=t[temp].f; if temp<>-1 then if t[temp].d then t[now].d:=true; end; for i:=1 to 4 do if t[now].n[i]<>0 then begin inc(tail); b[tail]:=t[now].n[i]; temp:=t[now].f; while (temp<>-1) and (t[temp].n[i]=0) do temp:=t[temp].f; if temp=-1 then t[t[now].n[i]].f:=0 else t[t[now].n[i]].f:=t[temp].n[i]; end else begin temp:=t[now].f; while (temp<>-1) and (t[temp].n[i]=0) do temp:=t[temp].f; if temp=-1 then t[now].n[i]:=0 else t[now].n[i]:=t[temp].n[i]; end; until head>=tail; end; procedure init; var i:longint; begin clr(0); t[0].f:=-1; tot:=0; readln(n); if n=0 then halt; for i:=1 to n do begin readln(a[i]); insert(a[i]); end; build_ac; end; function min(a,b:longint):longint; begin if a=-1 then a:=b else if (a>b) and (b<>-1) then a:=b; min:=a; end; procedure main; var i,j,k,len:longint; begin readln(s); len:=length(s); fillchar(f,sizeof(f),255); f[0,0]:=0; for i:=1 to len do for j:=0 to tot do if f[i-1,j]>-1 then for k:=1 to 4 do if not t[t[j].n[k]].d then f[i,t[j].n[k]]:=min(f[i,t[j].n[k]],f[i-1,j]+(1-ord(ch(s[i])=k))); ans:=-1; for i:=0 to tot do if not t[i].d then ans:=min(ans,f[len,i]); writeln(ans); end; begin while true do begin init; inc(k); write('Case k: '); main; end; end.