题目链接:http://acm.hdu.edu.cn/showproblem.php?pid=3567

Eight II

Time Limit: 4000/2000 MS (Java/Others)    Memory Limit: 130000/65536 K (Java/Others)
Total Submission(s): 3420    Accepted Submission(s): 742

Problem Description
Eight-puzzle, which is also called "Nine grids", comes from an old game.

In this game, you are given a 3 by 3 board and 8 tiles. The tiles are numbered from 1 to 8 and each covers a grid. As you see, there is a blank grid which can be represented as an 'X'. Tiles in grids having a common edge with the blank grid can be moved into
that blank grid. This operation leads to an exchange of 'X' with one tile.

We use the symbol 'r' to represent exchanging 'X' with the tile on its right side, and 'l' for the left side, 'u' for the one above it, 'd' for the one below it.

A state of the board can be represented by a string S using the rule showed below.

The problem is to operate an operation list of 'r', 'u', 'l', 'd' to turn the state of the board from state A to state B. You are required to find the result which meets the following constrains:
1. It is of minimum length among all possible solutions.
2. It is the lexicographically smallest one of all solutions of minimum length.

 
Input
The first line is T (T <= 200), which means the number of test cases of this problem.

The input of each test case consists of two lines with state A occupying the first line and state B on the second line.
It is guaranteed that there is an available solution from state A to B.

 
Output
For each test case two lines are expected.

The first line is in the format of "Case x: d", in which x is the case number counted from one, d is the minimum length of operation list you need to turn A to B.
S is the operation list meeting the constraints and it should be showed on the second line.

 
Sample Input
2
12X453786
12345678X
564178X23
7568X4123
 
Sample Output
Case 1: 2
dd
Case 2: 8
urrulldr
 
Author
zhymaoiing
 
Source

题解:

POJ1077 的强化版。

问:为什么加了vis判重比不加vis判重还要慢?

答:因为当引入vis判重时,就需要知道棋盘的状态,而计算一次棋盘的状态,就需要增加(8+7+……1)次操作,结果得不偿失。

更新:其实IDA*算法不能加vis判重,因为IDA*的本质就是dfs, 根据dfs的特性, 第一次被访问所用的步数并不一定是最少步数,所以如果加了vis判重,就默认取了第一次被访问时所用的步数,而这个步数不一定是最优的。所以第二份代码是错误的,即使过了oj的数据。

未加vis判重(202MS):

2017-09-10 10:25:57 Accepted 3567 202MS 1712K

 #include <iostream>
#include <cstdio>
#include <cstring>
#include <cmath>
#include <algorithm>
#include <vector>
#include <queue>
#include <stack>
#include <map>
#include <string>
#include <set>
#define ms(a,b) memset((a),(b),sizeof((a)))
using namespace std;
typedef long long LL;
const int INF = 2e9;
const LL LNF = 9e18;
const int MOD = 1e9+;
const int MAXN = 1e6+; //M为棋盘, pos_goal为目标状态的每个数字所在的位置, pos_goal[dig] = pos,
//即表明:在目标状态中,dig所在的位置为pos。pos_goal与M为两个互逆的数组。
int M[MAXN], pos_goal[MAXN]; int fac[] = { , , , , , , , , };
int dir[][] = { ,, ,-, ,, -, };
char op[] = {'d', 'l', 'r', 'u' }; int cantor(int s[]) //获得哈希函数值
{
int sum = ;
for(int i = ; i<; i++)
{
int num = ;
for(int j = i+; j<; j++)
if(s[j]<s[i]) num++;
sum += num*fac[-i];
}
return sum+;
} int dis_h(int s[]) //获得曼哈顿距离
{
int dis = ;
for(int i = ; i<; i++)
if(s[i]!=)
{
int x = i/, y = i%;
int xx = pos_goal[s[i]]/, yy = pos_goal[s[i]]%; //此处须注意
dis += abs(x-xx) + abs(y-yy);
}
return dis;
} char path[];
int kase, nextd;
bool IDAstar(int loc, int depth, int pre, int limit)
{
int h = dis_h(M);
if(depth+h>limit)
{
nextd = min(nextd, depth+h);
return false;
} if(h==)
{
path[depth] = '\0';
printf("Case %d: %d\n", kase, depth);
puts(path);
return true;
} int x = loc/;
int y = loc%;
for(int i = ; i<; i++)
{
if(i+pre==) continue; //方向与上一步相反, 剪枝
int xx = x + dir[i][];
int yy = y + dir[i][];
if(xx>= && xx<= && yy>= && yy<=)
{
int tmploc = xx*+yy;
swap(M[loc], M[tmploc]);
path[depth] = op[i];
if(IDAstar(xx*+yy, depth+, i, limit))
return true;
swap(M[loc], M[xx*+yy]);
}
}
return false;
} int main()
{
int T;
char str[];
scanf("%d",&T);
for(kase = ; kase<=T; kase++)
{
int loc;
scanf("%s", str);
for(int i = ; i<; i++)
{
if(str[i]=='X') M[i] = , loc = i;
else M[i] = str[i]-'';
} scanf("%s", str);
for(int i = ; i<; i++)
{
if(str[i]=='X') pos_goal[] = i;
else pos_goal[str[i]-''] = i;
} for(int limit = dis_h(M); ; limit = nextd) //迭代加深搜
{
nextd = INF;
if(IDAstar(loc, , INF, limit))
break;
}
}
}

加了vis判重(936MS)

2017-09-10 10:26:10 Accepted 3567 936MS 5620K

 #include <iostream>
#include <cstdio>
#include <cstring>
#include <cmath>
#include <algorithm>
#include <vector>
#include <queue>
#include <stack>
#include <map>
#include <string>
#include <set>
#define ms(a,b) memset((a),(b),sizeof((a)))
using namespace std;
typedef long long LL;
const int INF = 2e9;
const LL LNF = 9e18;
const int MOD = 1e9+;
const int MAXN = 1e6+; int M[MAXN], pos_goal[MAXN]; int fac[] = { , , , , , , , , };
int dir[][] = { ,, ,-, ,, -, };
char op[] = {'d', 'l', 'r', 'u' }; int cantor(int s[]) //获得哈希函数值
{
int sum = ;
for(int i = ; i<; i++)
{
int num = ;
for(int j = i+; j<; j++)
if(s[j]<s[i]) num++;
sum += num*fac[-i];
}
return sum+;
} int dis_h(int s[]) //获得曼哈顿距离
{
int dis = ;
for(int i = ; i<; i++)
if(s[i]!=)
{
int x = i/, y = i%;
int xx = pos_goal[s[i]]/, yy = pos_goal[s[i]]%;
dis += abs(x-xx) + abs(y-yy);
}
return dis;
} char path[];
int kase, nextd, vis[MAXN];
bool IDAstar(int loc, int depth, int pre, int limit)
{
int h = dis_h(M);
if(depth+h>limit)
{
nextd = min(nextd, depth+h);
return false;
} if(h==)
{
path[depth] = '\0';
printf("Case %d: %d\n", kase, depth);
puts(path);
return true;
} int x = loc/;
int y = loc%;
for(int i = ; i<; i++)
{
if(i+pre==) continue; //方向与上一步相反, 剪枝
int xx = x + dir[i][];
int yy = y + dir[i][];
if(xx>= && xx<= && yy>= && yy<=)
{
int tmploc = xx*+yy;
swap(M[loc], M[tmploc]);
int status = cantor(M);
if(!vis[status])
{
vis[status] = ;
path[depth] = op[i];
if(IDAstar(xx*+yy, depth+, i, limit))
return true;
vis[status] = ;
}
swap(M[loc], M[xx*+yy]);
}
}
return false;
} int main()
{
int T;
char str[];
scanf("%d",&T);
for(kase = ; kase<=T; kase++)
{
int loc;
scanf("%s", str);
for(int i = ; i<; i++)
{
if(str[i]=='X') M[i] = , loc = i;
else M[i] = str[i]-'';
} scanf("%s", str);
for(int i = ; i<; i++)
{
if(str[i]=='X') pos_goal[] = i;
else pos_goal[str[i]-''] = i;
} vis[cantor(M)] = ;
for(int limit = dis_h(M); ; limit = nextd) //迭代加深搜
{
nextd = INF;
ms(vis,);
if(IDAstar(loc, , INF, limit))
break;
}
}
}

HDU3567 Eight II —— IDA*算法的更多相关文章

  1. 【学时总结】 ◆学时·II◆ IDA*算法

    [学时·II] IDA*算法 ■基本策略■ 如果状态数量太多了,优先队列也难以承受:不妨再回头看DFS-- A*算法是BFS的升级,那么IDA*算法是对A*算法的再优化,同时也是对迭代加深搜索(IDF ...

  2. HUD 1043 Eight 八数码问题 A*算法 1667 The Rotation Game IDA*算法

    先是这周是搜索的题,网站:http://acm.hdu.edu.cn/webcontest/contest_show.php?cid=6041 主要内容是BFS,A*,IDA*,还有一道K短路的,.. ...

  3. LEETCODE —— Best Time to Buy and Sell Stock II [贪心算法]

    Best Time to Buy and Sell Stock II Say you have an array for which the ith element is the price of a ...

  4. HDU4513 吉哥系列故事——完美队形II Manacher算法

    题目链接:https://vjudge.net/problem/HDU-4513 吉哥系列故事——完美队形II Time Limit: 3000/1000 MS (Java/Others)    Me ...

  5. 八数码(IDA*算法)

    八数码 IDA*就是迭代加深和A*估价的结合 在迭代加深的过程中,用估计函数剪枝优化 并以比较优秀的顺序进行扩展,保证最早搜到最优解 需要空间比较小,有时跑得比A*还要快 #include<io ...

  6. HDU1560 DNA sequence —— IDA*算法

    题目链接:http://acm.hdu.edu.cn/showproblem.php?pid=1560 DNA sequence Time Limit: 15000/5000 MS (Java/Oth ...

  7. IDA*算法——骑士精神

    例题 骑士精神 Description 在一个5×5的棋盘上有12个白色的骑士和12个黑色的骑士, 且有一个空位.在任何时候一个骑士都能按照骑士的走法(它可以走到和它横坐标相差为1,纵坐标相差为2或者 ...

  8. UVA - 11212 Editing a Book(IDA*算法+状态空间搜索)

    题意:通过剪切粘贴操作,将n个自然段组成的文章,排列成1,2,……,n.剪贴板只有一个,问需要完成多少次剪切粘贴操作可以使文章自然段有序排列. 分析: 1.IDA*搜索:maxn是dfs的层数上限,若 ...

  9. 还不会ida*算法?看完这篇或许能理解点。

    IDA* 算法分析 IDA* 本质上就是带有估价函数和迭代加深优化的dfs与,A * 相似A *的本质便是带 有估价函数的bfs,估价函数是什么呢?估价函数顾名思义,就是估计由目前状态达 到目标状态的 ...

随机推荐

  1. 玩转css样式选择器----当父元素只有一个子元素时居中显示,多个水平排列

  2. C/C++怎样传递二维数组,转载自CSDN

    用二维数组作为参数传递(用二维数组处理矩阵),但是希望接受传递二维数组参数的函数可以处理任意维度的数组(希望矩阵的行数和列数都是不固定的). [以下转帖] ---------------------- ...

  3. 如何选择 IT 技术书籍

    ★第1招:看网上评论 首先,上一些权威的图书网站,看看大伙儿的评价如何(要相信群众的眼睛是雪亮的).对于英文书籍,我一般上亚马逊网站去看看:中文书籍则上豆瓣网.这两个网站都提供星级评分,一般 > ...

  4. HDU 5905 Black White Tree(树型DP)

    题目链接  Black White Tree 树型DP,设$f[i][j]$为以$i$为根的子树中大小为$j$的连通块中可以包含的最小黑点数目. $g[i][j]$为以$i$为根的子树中大小为$j$的 ...

  5. Go -- 一致性哈希算法

    一致性哈希算法在1997年由麻省理工学院的Karger等人在解决分布式Cache中提出的,设计目标是为了解决因特网中的热点(Hot spot)问题,初衷和CARP十分类似.一致性哈希修正了CARP使用 ...

  6. 文件重定向,getline()获取一样,屏幕输出流,格式控制符dec,oct,hex,精度控制setprecision(int num),设置填充,cout.width和file(字符),进制输入

     1.在window下的命令重定向输出到文件里 2.将内容输入到某个文件里的方式:命令<1.txt (使用1.txt中的命令) 3.读取文件里的名,然后将命令读取最后输出到文件里.命令< ...

  7. android 获取屏幕宽高 和 获取控件坐标

    一.获取屏幕宽高: (1). WindowManager wm = (WindowManager)getSystemService(Context.WINDOW_SERVICE); int width ...

  8. react 创建组件 (三)PureComponet

    我们知道,当组件的props或者state发生变化的时候:React会对组件当前的Props和State分别与nextProps和nextState进行比较,当发现变化时,就会对当前组件以及子组件进行 ...

  9. IE8.0登录Oracle EBS后报Oracle error 1403错

    IE8.0登录Oracle EBS后报错,登录页面打开没有问题.只是输入username和password然后登录,遇到下面错误: <PRE>Oracle error 1403: java ...

  10. Adobe Flash builder破解方法

    Flash Builder 4 有许多新的特性,可以结合新的功能使用新的Flex 4框架创建出更炫的应用.基于用户的反馈,对以数据中心的开发也进行了优化:对类如配置从服务器返回的数据类型这样的任务,也 ...