***uva 348 最优数组乘法序列（记忆化搜索+输出路径）

1、http://uva.onlinejudge.org/index.php?option=com_onlinejudge&Itemid=8&page=show_problem&problem=284

2、最优矩阵链乘，详见刘汝佳算法入门经典P170

3、题目：

 

Optimal Array Multiplication Sequence

Given two arrays A and B, we can determine the array C =A Busing the standard definition of matrix multiplication:

The number of columns in the A array must be the same as the number of rows in theB array. Notationally, let's say that rows(A) and columns(A) are the number of rows and columns, respectively, in theA array. The number of individual multiplications required to compute the entireC array (which will have the same number of rows as A and the same number of columns asB) is then rows(A) columns(B) columns(A). For example, ifA  is a    array, andB is a    array, it will take  , or 3000 multiplications to compute theC  array.

To perform multiplication of more than two arrays we have a choice of how to proceed. For example, ifX, Y, and Z are arrays, then to compute X Y Z we could either compute (X Y)Zor X (Y Z). Suppose X is a    array,Y is a    array, andZ is a    array. Let's look at the number of multiplications required to compute the product using the two different sequences:

(X Y) Z

•   multiplications to determine the product (X Y), a   array.
• Then    multiplications to determine the final result.
• Total multiplications: 4500.

X (Y Z)

•   multiplications to determine the product (Y Z), a   array.
• Then    multiplications to determine the final result.
• Total multiplications: 8750.

Clearly we'll be able to compute (X Y) Z using fewer individual multiplications.

Given the size of each array in a sequence of arrays to be multiplied, you are to determine an optimal computational sequence. Optimality, for this problem, is relative to the number of individual multiplications required.

Input

For each array in the multiple sequences of arrays to be multiplied you will be given only the dimensions of the array. Each sequence will consist of an integerN which indicates the number of arrays to be multiplied, and then N pairs of integers, each pair giving the number of rows and columns in an array; the order in which the dimensions are given is the same as the order in which the arrays are to be multiplied. A value of zero for N indicates the end of the input.N will be no larger than 10.

Output

Assume the arrays are named  . Your output for each input case is to be a line containing a parenthesized expression clearly indicating the order in which the arrays are to be multiplied. Prefix the output for each case with the case number (they are sequentially numbered, starting with 1). Your output should strongly resemble that shown in the samples shown below. If, by chance, there are multiple correct sequences, any of these will be accepted as a valid answer.

Sample Input

31 55 2020 135 1010 2020 35630 3535 1515 55 1010 2020 250

Sample Output

Case 1: (A1 x (A2 x A3))Case 2: ((A1 x A2) x A3)Case 3: ((A1 x (A2 x A3)) x ((A4 x A5) x A6))

 

4、AC代码：

#include<stdio.h>#include<string.h>#define N 15#define INF 0x7fffffffint dp[N][N];int path[N][N];struct node{    int x;    int y;}a[N];void Printpath(int i,int j){    if(i==j)    printf("A%d",i+1);    else    {        printf("(");        Printpath(i,path[i][j]);        printf(" x ");        Printpath(path[i][j]+1,j);        printf(")");    }}int DP(int i,int j){    if(i>=j)    return 0;    if(dp[i][j]!=0)    return dp[i][j];    dp[i][j]=INF;    for(int k=i;k<j;k++)    {        int tmp=DP(i,k)+DP(k+1,j)+a[i].x*a[k].y*a[j].y;        if(tmp<dp[i][j])        {            dp[i][j]=tmp;            path[i][j]=k;        }    }    return dp[i][j];}int main(){    int n,cas=0;    while(scanf("%d",&n)!=EOF)    {        cas++;        if(n==0)        break;        for(int i=0;i<n;i++)        scanf("%d%d",&a[i].x,&a[i].y);        memset(dp,0,sizeof(dp));        memset(path,0,sizeof(path));        int res=DP(0,n-1);        printf("Case %d: ",cas);        Printpath(0,n-1);        printf("\n");    }    return 0;}/*31 55 2020 135 1010 2020 35630 3535 1515 55 1010 2020 250*/