回溯法是基本算法的一种,可以用于解决大致这样的问题:假设我们有一个N个元素的集合{N},现在要依据该集合生成M个元素的集合{M},每一个元素的生成都依据一定的规则CHECK。
用回溯法解决此问题,我们可以划分为三个重要组成部分。
步骤
从第一步开始至第M步,每一步都从{N}中选取一个元素放入结果{M}中。
界定
每次选择一个元素时,我们都要用规则CHECK来界定{N}中的元素谁合适。界定规则的描述将决定算法的效率和性能。
回溯
如果第k步不能找到合适的元素或者需要得到更多的结果,返回到第k-1步,继续选择下一个第k-1步的元素。
让我们来运用以上的描述和C++语言来解决数学排列和组合问题。
问题1:从N个元素中选择M个元素进行排列,列出所有结果。
问题2:从N个元素中选择M个元素进行组合,列出所有结果。
与排列不同的是,组合不需要对选择出来的M个元素进行排列,不妨假定每一组结果中的M个元素从小到大排列。
当然,如果将回溯法运用到工程中去,正确性只是必要条件之一,效率显得相当重要,高效的界定代码则是其中的关键。甚至可以考虑换一种方法来实现,当然回溯的思想应该都一样。下面一段代码实现N的阶乘(排列的一个特例),是我在工程中的一个应用。思路则是假设我们已经按照一种方式列出了所有结果
1, 2, 3
1, 3, 2
2, 1, 3
2, 3, 1
3, 1, 2
3, 2, 1
然后纵向的填写所有结果至结果数组中。
事实上,如果你在工程应用中用到的只是可列举的排列组合结果,你应该考虑把结果直接编码到程序中去,这样最快了,只是扩展性不好罢了。
用回溯法解决此问题,我们可以划分为三个重要组成部分。
步骤
从第一步开始至第M步,每一步都从{N}中选取一个元素放入结果{M}中。
界定
每次选择一个元素时,我们都要用规则CHECK来界定{N}中的元素谁合适。界定规则的描述将决定算法的效率和性能。
回溯
如果第k步不能找到合适的元素或者需要得到更多的结果,返回到第k-1步,继续选择下一个第k-1步的元素。
让我们来运用以上的描述和C++语言来解决数学排列和组合问题。
问题1:从N个元素中选择M个元素进行排列,列出所有结果。
//
permutation.h : List all the permutation subsets of a certian set.
//
#pragma once
#include < iostream >
#include < iomanip >
using namespace std;
// Compute P(N, M). N is the total number, M is the selected number.
template < int N, int M >
class CPermutation
{
private:
int** m_result; // two-dimension array of [m_nCount][M]
int m_nCount; // how many results
int m_nIndex; // 0 - m_nCount - 1
public:
// List all possible results by nesting
void Count()
{
m_nIndex = 0;
int result[N], used[N];
for (int i = 0; i < N; i++)
used[i] = 0;
CountRecur(result, used, 0);
}
void CountRecur(int result[M], int used[M], int i)
{
for (int k = 0; k < N; k++)
{
if (used[k])
continue ;
result[i] = k;
used[k] = 1;
if (i < M - 1)
CountRecur(result, used, i + 1);
else
Add(result);
used[k] = 0;
}
}
// Save the result
void Add(int sz[M])
{
memcpy(m_result[m_nIndex], sz, M * sizeof(int));
++m_nIndex;
}
// Count the number of subsets
// C(N, M) = N! / ((N - M)! * M!)
static int NumberOfResult()
{
if (N <= 0 || M <= 0 || M > N)
return 0;
int result = 1;
for (int i = 0; i < M; i++)
result *= N - i;
return result;
}
// Print them to the standard output device
void Print()
{
for (int i = 0; i < m_nCount; i++)
{
cout << setw(3) << setfill(' ') << i + 1 << ":";
for (int j = 0; j < M; j++)
cout << setw(3) << setfill(' ') << m_result[i][j] + 1;
cout << endl;
}
}
CPermutation()
{
// allocate memories for the result
m_nCount = NumberOfResult();
m_result = new int*[m_nCount];
for (int i = 0; i < m_nCount; i++)
m_result[i] = new int[M];
}
~CPermutation()
{
// deallocate memories for the result
for (int i = 0; i < m_nCount; i++)
delete[] m_result[i];
delete[] m_result;
}
} ;
//
#pragma once
#include < iostream >
#include < iomanip >
using namespace std;
// Compute P(N, M). N is the total number, M is the selected number.
template < int N, int M >
class CPermutation
{
private:
int** m_result; // two-dimension array of [m_nCount][M]
int m_nCount; // how many results
int m_nIndex; // 0 - m_nCount - 1
public:
// List all possible results by nesting
void Count()
{
m_nIndex = 0;
int result[N], used[N];
for (int i = 0; i < N; i++)
used[i] = 0;
CountRecur(result, used, 0);
}
void CountRecur(int result[M], int used[M], int i)
{
for (int k = 0; k < N; k++)
{
if (used[k])
continue ;
result[i] = k;
used[k] = 1;
if (i < M - 1)
CountRecur(result, used, i + 1);
else
Add(result);
used[k] = 0;
}
}
// Save the result
void Add(int sz[M])
{
memcpy(m_result[m_nIndex], sz, M * sizeof(int));
++m_nIndex;
}
// Count the number of subsets
// C(N, M) = N! / ((N - M)! * M!)
static int NumberOfResult()
{
if (N <= 0 || M <= 0 || M > N)
return 0;
int result = 1;
for (int i = 0; i < M; i++)
result *= N - i;
return result;
}
// Print them to the standard output device
void Print()
{
for (int i = 0; i < m_nCount; i++)
{
cout << setw(3) << setfill(' ') << i + 1 << ":";
for (int j = 0; j < M; j++)
cout << setw(3) << setfill(' ') << m_result[i][j] + 1;
cout << endl;
}
}
CPermutation()
{
// allocate memories for the result
m_nCount = NumberOfResult();
m_result = new int*[m_nCount];
for (int i = 0; i < m_nCount; i++)
m_result[i] = new int[M];
}
~CPermutation()
{
// deallocate memories for the result
for (int i = 0; i < m_nCount; i++)
delete[] m_result[i];
delete[] m_result;
}
} ;
问题2:从N个元素中选择M个元素进行组合,列出所有结果。
与排列不同的是,组合不需要对选择出来的M个元素进行排列,不妨假定每一组结果中的M个元素从小到大排列。
//
combination.h : list all the subsets of a certian set
//
#pragma once
#include < iostream >
#include < iomanip >
using namespace std;
// List all the M-subsets of the N-set
template < int N, int M >
class CCombination
{
private:
int** m_result; // two-dimension array of m_nCount * M
int m_nCount; // how many results of M-length array
public:
CCombination()
{
// allocate memories for the result
m_nCount = NumberOfResult();
m_result = new int*[m_nCount];
for (int i = 0; i < m_nCount; i++)
m_result[i] = new int[M];
}
~CCombination()
{
// deallocate memories for the result
for (int i = 0; i < m_nCount; i++)
delete[] m_result[i];
delete[] m_result;
}
// process of counting
void Count()
{
int sz[M];
int nResultCount = 0;
CountRecur(sz, 0, 0, nResultCount);
}
// Print them to the standard output device
void Print()
{
using std::cout;
using std::setw;
using std::setfill;
using std::endl;
for (int i = 0; i < m_nCount; i++)
{
cout << setw(3) << setfill(' ') << i + 1 << ":";
for (int j = 0; j < M; j++)
cout << setw(3) << setfill(' ') << m_result[i][j] + 1;
cout << endl;
}
}
private:
// Count the number of subsets
// C(N, M) = N! / ((N - M)! * M!)
int NumberOfResult()
{
int result = 1;
for (int i = 0; i < M; i++)
result *= N - i;
for (int j = 1; j <= M; j++)
result /= j;
return result;
}
// Get the current value
// sz - array of the current result
// nIndex - index of sz, 0 <= nIndex < M
// nStartVal - the current minimum value, 0 <= nStartVal < N
// nResultCount - index of m_result
void CountRecur(int sz[M], int nIndex, int nStartVal, int& nResultCount)
{
if (nStartVal + M - nIndex > N)
return ;
for (int i = nStartVal; i < N; i++)
{
sz[nIndex] = i;
if (nIndex == M - 1)
Add(sz, nResultCount);
else
CountRecur(sz, nIndex + 1, i + 1, nResultCount);
}
}
// Save the result
void Add(int* sz, int& nIndex)
{
memcpy(m_result[nIndex], sz, M * sizeof(int));
++nIndex;
}
} ;
//
#pragma once
#include < iostream >
#include < iomanip >
using namespace std;
// List all the M-subsets of the N-set
template < int N, int M >
class CCombination
{
private:
int** m_result; // two-dimension array of m_nCount * M
int m_nCount; // how many results of M-length array
public:
CCombination()
{
// allocate memories for the result
m_nCount = NumberOfResult();
m_result = new int*[m_nCount];
for (int i = 0; i < m_nCount; i++)
m_result[i] = new int[M];
}
~CCombination()
{
// deallocate memories for the result
for (int i = 0; i < m_nCount; i++)
delete[] m_result[i];
delete[] m_result;
}
// process of counting
void Count()
{
int sz[M];
int nResultCount = 0;
CountRecur(sz, 0, 0, nResultCount);
}
// Print them to the standard output device
void Print()
{
using std::cout;
using std::setw;
using std::setfill;
using std::endl;
for (int i = 0; i < m_nCount; i++)
{
cout << setw(3) << setfill(' ') << i + 1 << ":";
for (int j = 0; j < M; j++)
cout << setw(3) << setfill(' ') << m_result[i][j] + 1;
cout << endl;
}
}
private:
// Count the number of subsets
// C(N, M) = N! / ((N - M)! * M!)
int NumberOfResult()
{
int result = 1;
for (int i = 0; i < M; i++)
result *= N - i;
for (int j = 1; j <= M; j++)
result /= j;
return result;
}
// Get the current value
// sz - array of the current result
// nIndex - index of sz, 0 <= nIndex < M
// nStartVal - the current minimum value, 0 <= nStartVal < N
// nResultCount - index of m_result
void CountRecur(int sz[M], int nIndex, int nStartVal, int& nResultCount)
{
if (nStartVal + M - nIndex > N)
return ;
for (int i = nStartVal; i < N; i++)
{
sz[nIndex] = i;
if (nIndex == M - 1)
Add(sz, nResultCount);
else
CountRecur(sz, nIndex + 1, i + 1, nResultCount);
}
}
// Save the result
void Add(int* sz, int& nIndex)
{
memcpy(m_result[nIndex], sz, M * sizeof(int));
++nIndex;
}
} ;
当然,如果将回溯法运用到工程中去,正确性只是必要条件之一,效率显得相当重要,高效的界定代码则是其中的关键。甚至可以考虑换一种方法来实现,当然回溯的思想应该都一样。下面一段代码实现N的阶乘(排列的一个特例),是我在工程中的一个应用。思路则是假设我们已经按照一种方式列出了所有结果
1, 2, 3
1, 3, 2
2, 1, 3
2, 3, 1
3, 1, 2
3, 2, 1
然后纵向的填写所有结果至结果数组中。
//
pi : this algorith lists all the sequences of a certian array
//
#pragma once
#include " common.h "
// how many sequence for N cells
int NumberOfPI( int N)
{
if (N < 0)
return 0;
if (N == 0)
return 1;
int pi = 1;
for (int i = 1; i <= N; i++)
pi *= i;
return pi;
}
// print all the sequence to a certain array
// N - number of cells
// PI - array of cells
// selPI - index of array with initial value of 0 and maximum value of N - 1
// Index - result array
// selIndex - number of result array
void PrintPI( int N, int * PI, int selPI, int ** Index, int selIndex)
{
if (N <= 0)
return ;
int num = NumberOfPI(N - selPI - 1);
for (int i = selPI; i < N; i++)
{
swap(PI[selPI], PI[i]);
PrintPI(N, PI, selPI + 1, Index, selIndex);
for (int j = 0; j < num; j++)
Index[selIndex++][selPI] = PI[selPI];
swap(PI[selPI], PI[i]);
}
}
int PrintPITest()
{
// Number
static const int N = 5;
// Result in all sequences
int PI[N];
for (int l = 0; l < N; l++)
PI[l] = l + 1;
// Allocate the two_demension array
const int num = NumberOfPI(N);
int **Index;
Index = new int*[num];
for (int l = 0; l < num; l++)
Index[l] = new int[N];
// Calculating.
PrintPI(N, PI, 0, Index, 0);
// Print it to cout
for (int i = 0; i < num; i++)
{
printf("%5d : ", i + 1);
for (int j = 0; j < N; j++)
printf("%2d ", Index[i][j]);
printf("\n");
}
// Check if there is any identical indexes.
for (int i = 0; i < num - 1; i++)
for (int j = i + 1; j < num; j++)
{
int k = 0;
for (; k < N; k++)
if (Index[i][k] != Index[j][k])
break;
if (k == N)
printf("Check ERROR!\n");
}
// Release the two_demension array
for (int l = 0; l < num; l++)
delete[] Index[l];
delete[] Index;
return 0;
}
//
#pragma once
#include " common.h "
// how many sequence for N cells
int NumberOfPI( int N)
{
if (N < 0)
return 0;
if (N == 0)
return 1;
int pi = 1;
for (int i = 1; i <= N; i++)
pi *= i;
return pi;
}
// print all the sequence to a certain array
// N - number of cells
// PI - array of cells
// selPI - index of array with initial value of 0 and maximum value of N - 1
// Index - result array
// selIndex - number of result array
void PrintPI( int N, int * PI, int selPI, int ** Index, int selIndex)
{
if (N <= 0)
return ;
int num = NumberOfPI(N - selPI - 1);
for (int i = selPI; i < N; i++)
{
swap(PI[selPI], PI[i]);
PrintPI(N, PI, selPI + 1, Index, selIndex);
for (int j = 0; j < num; j++)
Index[selIndex++][selPI] = PI[selPI];
swap(PI[selPI], PI[i]);
}
}
int PrintPITest()
{
// Number
static const int N = 5;
// Result in all sequences
int PI[N];
for (int l = 0; l < N; l++)
PI[l] = l + 1;
// Allocate the two_demension array
const int num = NumberOfPI(N);
int **Index;
Index = new int*[num];
for (int l = 0; l < num; l++)
Index[l] = new int[N];
// Calculating.
PrintPI(N, PI, 0, Index, 0);
// Print it to cout
for (int i = 0; i < num; i++)
{
printf("%5d : ", i + 1);
for (int j = 0; j < N; j++)
printf("%2d ", Index[i][j]);
printf("\n");
}
// Check if there is any identical indexes.
for (int i = 0; i < num - 1; i++)
for (int j = i + 1; j < num; j++)
{
int k = 0;
for (; k < N; k++)
if (Index[i][k] != Index[j][k])
break;
if (k == N)
printf("Check ERROR!\n");
}
// Release the two_demension array
for (int l = 0; l < num; l++)
delete[] Index[l];
delete[] Index;
return 0;
}
事实上,如果你在工程应用中用到的只是可列举的排列组合结果,你应该考虑把结果直接编码到程序中去,这样最快了,只是扩展性不好罢了。