矩阵连乘问题
【问题】
给定n个矩阵的链<A1,A2,…,An>,其中Ai与是Ai-1可乘的,矩阵Ai的维数为pi-1*pi(1≤i≤n),
如何确定计算矩阵链乘积A1A2…An的计算次序(完全括号化方式),使得依此次序计算矩阵链乘积需要的数乘次数最少。
【算法分析】
将矩阵连乘积AiAi-1...Aj简记为A[i:j] ,i≤j 。
考察计算A[i:j]的最优计算次序。设这个计算次序在矩阵Ak和Ak+1之间将矩阵链断开,i≤k<j,则其相应完全加括号方式为:
(AiAi+1...Ak)(Ak+1Ak+2...Aj)
计算量:A[i:k]的计算量加上A[k+1:j]的计算量,再加上A[i:k]和A[k+1:j]相乘的计算量。
特征:计算A[i:j]的最优次序所包含的计算矩阵子链 A[i:k]和A[k+1:j]的次序也是最优的。
矩阵连乘计算次序问题的最优解包含着其子问题的最优解,这种性质称为最优子结构性质。
递归式如下:
【源代码】
代码(1)
#include <iostream> using namespace std; void MatrixChain(int p[], int **m, int **s, int n) { //将对角线元素赋值为零,即单个矩阵计算量为0 for (int i = 1; i <= n; i++) m[i][i] = 0; //r是矩阵相乘个数 for (int r = 2; r <= n; r++) for (int i = 1; i <= n - r+1; i++) { int j = i + r - 1; //计算A[i, j] = A[i: i] A[i+1: j] m[i][j] = m[i + 1][j] + p[i-1] * p[i] * p[j]; s[i][j] = i; //记下断点i for (int k = i + 1; k < j; k++) { //对i<k<j, 逐个计算A[i, j] = A[i: k] A[k+1: j] int t = m[i][k] + m[k + 1][j] + p[i-1] * p[k] * p[j]; //记下较小的m[i][j]及相应的断点k if (t < m[i][j]) { m[i][j] = t; s[i][j] = k; } } } } //找出s数组中记录的最优断开点 void Traceback(int i, int j, int **s) { if (i == j) { cout << "A" << i; return; } cout << "("; Traceback(i, s[i][j], s); Traceback(s[i][j] + 1, j, s); cout << ")"; } int main() { const int n = 6; int p[n + 1] = {30,35,15,5,10,20,25 }; int **m = new int*[n+1]; int **s = new int*[n+1]; for (int i = 0; i < n+1; i++) { m[i] = new int[n+1]; s[i] = new int[n+1]; } MatrixChain(p, m, s, n); Traceback(1, n, s); for (int i = 0; i < n+1; i++) { delete[] m[i]; delete[] s[i]; } delete[]m; delete[]s; return 0; }
代码(2)
#include <iostream> using namespace std; void MatrixChain(int p[], int **m, int **s, int n) { //将对角线元素赋值为零,即单个矩阵计算量为0 for (int i = 0; i < n; i++) m[i][i] = 0; //r是矩阵相乘个数 for (int r = 2; r <= n; r++) for (int i = 0; i <= n - r; i++) { int j = i + r - 1; //计算A[i, j] = A[i: i] A[i+1: j] m[i][j] = m[i + 1][j] + p[i] * p[i + 1] * p[j + 1]; s[i][j] = i; //记下断点i for (int k = i + 1; k < j; k++) { //对i<k<j, 逐个计算A[i, j] = A[i: k] A[k+1: j] int t = m[i][k] + m[k + 1][j] + p[i] * p[k + 1] * p[j + 1]; //记下较小的m[i][j]及相应的断点k if (t < m[i][j]) { m[i][j] = t; s[i][j] = k; } } } } //找出s数组中记录的最优断开点 void Traceback(int i, int j, int **s) { if (i == j) { cout << "A" << i + 1; return; } cout << "("; Traceback(i, s[i][j], s); Traceback(s[i][j] + 1, j, s); cout << ")"; } int main() { const int n = 6; int p[n + 1] = {30,35,15,5,10,20,25 }; int **m = new int*[n]; int **s = new int*[n]; for (int i = 0; i < n; i++) { m[i] = new int[n]; s[i] = new int[n]; } MatrixChain(p, m, s, n); Traceback(0, n - 1, s); for (int i = 0; i < n; i++) { delete[] m[i]; delete[] s[i]; } delete[]m; delete[]s; return 0; }
【算法复杂度分析】
主要计算量取决于算法中对r,i和k的3重循环。
算法的计算时间上界为O(n3)。
算法所占用的空间为O(n2)。