欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页  >  IT编程

【算法导论】--分治策略Strassen算法(运用下标运算)【c++】

程序员文章站 2023-02-02 10:29:23
由于偷懒不想用泛型,所以直接用了整型来写了一份 ①首先你得有一个矩阵的class Matrix ②Matrix为了方便用下标进行运算, Matrix的结构如图:(我知道我的字丑。。。) Matrix.h代码如下:(个人并不喜欢把代码全写在一块,对于阅读者是相当巨大的负担,其实自己受不了(逃)) Ma ......

由于偷懒不想用泛型,所以直接用了整型来写了一份

①首先你得有一个矩阵的class matrix

②matrix为了方便用下标进行运算,

matrix的结构如图:(我知道我的字丑。。。)

【算法导论】--分治策略Strassen算法(运用下标运算)【c++】

 

 matrix.h代码如下:(个人并不喜欢把代码全写在一块,对于阅读者是相当巨大的负担,其实自己受不了(逃))

 1 #pragma once
 2 #include<vector>
 3 using namespace std;
 4 class matrix
 5 {
 6 public:
 7     vector<vector<int>> nums;
 8     int x0, y0;
 9 
10     int size;
11     matrix();
12     ~matrix();
13     matrix(int size, int x0, int y0);
14     matrix(matrix input, int x0, int y0, int size);
15     matrix(vector<vector<int>>input);
16     void display();
17     static void matrixmultiinit(matrix &a, matrix &b);
18     static matrix matrixmulti(matrix a, matrix b);
19     static matrix matrixadd(matrix a, matrix b);
20     static matrix matrixsub(matrix a, matrix b);
21 };

 

 

matrix.cpp对类的实现

一,构造,析构函数

 1 matrix::matrix()
 2 {
 3 }
 4 
 5 
 6 matrix::~matrix()
 7 {
 8 }
 9 
10 matrix::matrix(int size, int x0, int y0)
11 {
12     vector<vector<int>> temp(size, *new vector<int>(size));
13     for (int i = 0; i < size; i++)
14         for (int j = 0; j < size; j++)
15             temp[i][j] = 0;
16     this->nums = temp;
17     this->x0 = x0;
18     this->y0 = y0;
19     this->size = size;
20 
21 }
22 
23 matrix::matrix(matrix input, int x0, int y0, int size)
24 {
25     this->nums = input.nums;
26     this->x0 = x0;
27     this->y0 = y0;
28     this->size = size;
29 }
30 
31 matrix::matrix(vector<vector<int>> input)
32 {
33     this->nums = input;
34     this->x0 = 0;
35     this->y0 = 0;
36     this->size = input.size();
37 }

 

二,a+b,a-b实现

 1 matrix matrix::matrixadd(matrix a, matrix b)
 2 {
 3     matrix result(a.size, 0, 0);
 4     for (int i = 0; i < result.nums.size(); i++)
 5         for (int j = 0; j < result.nums.size(); j++)
 6             result.nums[i][j] = a.nums[a.x0 + i][a.y0 + j] + b.nums[b.x0 + i][b.y0 + j];
 7     return result;
 8 }
 9 
10 matrix matrix::matrixsub(matrix a, matrix b)
11 {
12 
13     matrix result(a.size, 0, 0);
14     for (int i = 0; i < result.nums.size(); i++)
15         for (int j = 0; j < result.nums.size(); j++)
16             result.nums[i][j] = a.nums[a.x0 + i][a.y0 + j] - b.nums[b.x0 + i][b.y0 + j];
17     return result;
18 }

 

三,a*b的实现

 1 matrix matrix::matrixmulti(matrix a, matrix b)
 2 {
 3     int n = a.size;
 4     int halfsize =n / 2;
 5     matrix result(n, 0, 0);
 6     if (n == 1)
 7         result.nums[0][0] = a.nums[a.x0][a.y0] * b.nums[b.x0][b.y0];
 8     else
 9     {
10         matrix temps[10];
11         for (int i = 0; i < 10; i++)
12         {
13             temps[i] = *new matrix(halfsize, 0, 0);
14         }
15 
16         //00--  a.x0,a.y0,halfsize
17         //01-- a.x0,a.y0+halfsize,halfsize
18         //10--  a.x0+halfsize,a.y0,halfsize
19         //11--    a.x0+halfsize,a.y0+halfsize,halfsize
20                 
21         //00-- b.x0,b.y0,halfsize
22         //01-- b.x0,b.y0+halfsize,halfsize
23         //10-- b.x0+halfsize,b.y0,halfsize
24         //11--    b.x0+halfsize,b.y0+halfsize,halfsize
25         temps[0] = temps[0].matrixsub(*new matrix(b, b.x0, b.y0 + halfsize, halfsize), *new matrix(b, b.x0 + halfsize, b.y0 + halfsize, halfsize));//01-11 b
26         temps[1] = temps[1].matrixadd(*new matrix(a, a.x0, a.y0, halfsize), *new matrix(a, a.x0, a.y0 + halfsize, halfsize));//00+01 a
27         temps[2] = temps[2].matrixadd(*new matrix(a, a.x0 + halfsize, a.y0, halfsize), *new matrix(a, a.x0 + halfsize, a.y0 + halfsize, halfsize));//10+11 a
28         temps[3] = temps[3].matrixsub(*new matrix(b, b.x0 + halfsize, b.y0, halfsize), *new matrix(b, b.x0, b.y0, halfsize));//10-00 b
29         temps[4] = temps[4].matrixadd(*new matrix(a, a.x0, a.y0, halfsize), *new matrix(a, a.x0 + halfsize, a.y0 + halfsize, halfsize));//00+11 a
30         temps[5] = temps[5].matrixadd(*new matrix(b, b.x0, b.y0, halfsize), *new matrix(b, b.x0 + halfsize, b.y0 + halfsize, halfsize));//00+11 b
31         temps[6] = temps[6].matrixsub(*new matrix(a, a.x0, a.y0 + halfsize, halfsize), *new matrix(a, a.x0 + halfsize, a.y0 + halfsize, halfsize));//01-11 a
32         temps[7] = temps[7].matrixadd(*new matrix(b, b.x0 + halfsize, b.y0, halfsize), *new matrix(b, b.x0 + halfsize, b.y0 + halfsize, halfsize));//10+11 b
33         temps[8] = temps[8].matrixsub(*new matrix(a, a.x0, a.y0, halfsize), *new matrix(a, a.x0 + halfsize, a.y0, halfsize));//00-10 a
34         temps[9] = temps[9].matrixadd(*new matrix(b, b.x0, b.y0, halfsize), *new matrix(b, b.x0, b.y0 + halfsize, halfsize));//00+01 b
35 
36 
37         matrix tempp[7];
38         for (int i = 0; i < 7; i++)
39         {
40             tempp[i] = *new matrix(n / 2, 0, 0);
41         }
42         tempp[0] = tempp[0].matrixmulti(*new matrix(a, a.x0, a.y0, halfsize), *new matrix(temps[0], 0, 0, halfsize));
43         tempp[1] = tempp[1].matrixmulti(*new matrix(temps[1], 0, 0,halfsize), *new matrix(b, b.x0 + halfsize, b.y0 + halfsize, halfsize));
44         tempp[2] = tempp[2].matrixmulti(*new matrix(temps[2], 0, 0, halfsize), *new matrix(b, b.x0, b.y0, halfsize));
45         tempp[3] = tempp[3].matrixmulti(*new matrix(a, a.x0 + halfsize, a.y0 + halfsize, halfsize), *new matrix(temps[3], 0, 0, halfsize));
46         tempp[4] = tempp[4].matrixmulti(*new matrix(temps[4], 0, 0, halfsize), *new matrix(temps[5], 0, 0, halfsize));
47         tempp[5] = tempp[5].matrixmulti(*new matrix(temps[6], 0, 0, halfsize), *new matrix(temps[7], 0, 0, halfsize));
48         tempp[6] = tempp[6].matrixmulti(*new matrix(temps[8], 0, 0, halfsize), *new matrix(temps[9], 0, 0, halfsize));
49 
50 
51 
52         matrix result00 = result00.matrixadd(tempp[4], tempp[3]);
53         result00 = result00.matrixsub(result00, tempp[1]);
54         result00 = result00.matrixadd(result00, tempp[5]);
55         matrix result01 = result01.matrixadd(tempp[0], tempp[1]);
56         matrix result10 = result10.matrixadd(tempp[2], tempp[3]);
57         matrix result11 = result11.matrixadd(tempp[4], tempp[0]);
58         result11 = result11.matrixsub(result11, tempp[2]);
59         result11 = result11.matrixsub(result11, tempp[6]);
60 
61         if (n == 3) {
62         for(int i=0;i<n/2+1;i++)
63             for (int j = 0; j < n / 2 + 1; j++) {
64             
65             result.nums[i][j]= result00.nums[i][j];
66             result.nums[i][j+n/2+1] = result01.nums[i][j];
67             result.nums[i+n/2+1][j] = result10.nums[i][j];
68             result.nums[i+n/2+1][j+n/2+1] = result11.nums[i][j];
69             }
70         }
71 
72         for(int i=0;i<n/2;i++)
73             for (int j = 0; j < n / 2; j++) {
74             
75             result.nums[i][j]= result00.nums[i][j];
76             result.nums[i][j+n/2] = result01.nums[i][j];
77             result.nums[i+n/2][j] = result10.nums[i][j];
78             result.nums[i+n/2][j+n/2] = result11.nums[i][j];
79             }
80 
81     }
82     return result;
83 }

 

四,防止size%2!=0的处理函数(即矩阵的行列数为奇数)

 1 void matrix::matrixmultiinit(matrix &a, matrix &b) {
 2 
 3     if (a.nums.size() % 2 != 0)
 4     {
 5         for (int i = 0; i < a.nums.size(); i++)
 6             a.nums[i].push_back(0);
 7         for (int i = 0; i < b.nums.size(); i++)
 8             b.nums[i].push_back(0);
 9         a.nums.push_back(*new vector<int>(a.nums[0].size(), 0));
10         b.nums.push_back(*new vector<int>(b.nums[0].size(), 0));
11         a.size++;
12         b.size++;
13     }
14 }

 

五,输出函数(这个读者随意)

 1 void matrix::display()
 2 {
 3     for (int i = 0; i < this->nums.size(); i++) {
 4 
 5         cout << "||";
 6         for (int j = 0; j < this->nums[i].size(); j++) {
 7             cout << this->nums[i][j] << " ";
 8         }
 9         cout << "||" << endl;
10 
11     }
12 }

 

六,测试函数

 1 #include <iostream>
 2 #include"matrix.h"
 3 int main()
 4 {
 5     vector<vector<int>> input = { 
 6         {1,2,3},
 7         {1,2,3},
 8         {1,1,1},
 9     };
10     matrix test0 = * new matrix(input);
11     matrix test1 = * new matrix(input);
12     matrix test2;
13     test2.matrixmultiinit(test0, test1);
14     test2= test2.matrixmulti(test0, test1);
15     test2.display();
16 
17 }

 

本人比较愚笨,耗时一天半才完成,不知道是不是天气热的原因,人太燥了,沉不下心来思考bug。

 

a*b有一点需要注意的是分块的逻辑应该怎么表示,一开始我用了两个顶点来表示一个矩阵的分块,如图:

【算法导论】--分治策略Strassen算法(运用下标运算)【c++】

然后halfsize还得一个一个的算,然后自己敲错的几率还会加大,并且还不一定表示正确每个分块,然后就逼疯自己了。

感觉这两天被这一堆bug都弄自闭了。。。。

幸好还是撑过来了,这算是我啃算法导论的第一个坎吧。

幸好第二天在csdn里面看到了别人怎么分块的。看到了一个变量halfsize,于是才开始改自己matrix的构造。

虽然一开始不愿意,但是改完之后,竟然一次过了!woc!

给我了一个教训,以后能“少”一个变量尽量“少”一个变量。