C#0-1背包问题已知总和和集合中的零数

时间:2011-09-04 08:09:41

标签: c# algorithm knapsack-problem

我有一个5x5的值,从0到3(含),所有值都未知。我知道每个行和列的值的总和和零的数量。我将如何使用C#解决这个0-1背包问题,并检索满足已知总和和零数的可能解决方案?这些表总是五行五列,所以它不是一个传统的背包。

例如,我们输入:

Row[0]: Sum=4, Zeros=1
   [1]: Sum=5, Zeros=1
   [2]: Sum=4, Zeros=2
   [3]: Sum=8, Zeros=0
   [4]: Sum=3, Zeros=2

Col[0]: Sum=5, Zeros=1
   [1]: Sum=3, Zeros=2
   [2]: Sum=4, Zeros=2
   [3]: Sum=5, Zeros=1
   [4]: Sum=7, Zeros=0

我们会将此作为一种可能的解决方案:

[[ 0 1 1 1 1 ]
 [ 1 0 2 1 1 ]
 [ 2 1 0 0 1 ]
 [ 1 1 1 2 3 ]
 [ 1 0 0 1 1 ]]

在这种相当奇怪的情况下,我应该采用什么类型的算法?我是否还必须编写一个类来枚举排列?

编辑以澄清:问题不在于我无法列举可能性;我不知道如何在包含指定数量的零和最多5个项目的同时有效地确定添加到任意总和的排列。

2 个答案:

答案 0 :(得分:3)

这里有代码。如果您需要任何评论,请随时询问:

using System;
using System.Diagnostics;

namespace ConsoleApplication15
{
    class Program
    {
        static void Main(string[] args)
        {
            RowOrCol[] rows = new RowOrCol[] { 
                new RowOrCol(4, 1),
                new RowOrCol(5, 1),
                new RowOrCol(4, 2),
                new RowOrCol(8, 0),
                new RowOrCol(3, 2),
            };

            RowOrCol[] cols = new RowOrCol[] { 
                new RowOrCol(5, 1),
                new RowOrCol(3, 2),
                new RowOrCol(4, 2),
                new RowOrCol(5, 1),
                new RowOrCol(7, 0),
            };

            int[,] table = new int[5, 5];

            Stopwatch sw = Stopwatch.StartNew();

            int solutions = Do(table, rows, cols, 0, 0);

            sw.Stop();

            Console.WriteLine();
            Console.WriteLine("Found {0} solutions in {1}ms", solutions, sw.ElapsedMilliseconds);
            Console.ReadKey();
        }

        public static int Do(int[,] table, RowOrCol[] rows, RowOrCol[] cols, int row, int col)
        {
            int solutions = 0;

            int oldValueRowSum = rows[row].Sum;
            int oldValueRowZero = rows[row].Zeros;
            int oldValueColSum = cols[col].Sum;
            int oldValueColZero = cols[col].Zeros;

            int nextCol = col + 1;
            int nextRow;
            bool last = false;

            if (nextCol == cols.Length)
            {
                nextCol = 0;

                nextRow = row + 1;

                if (nextRow == rows.Length)
                {
                    last = true;
                }
            }
            else
            {
                nextRow = row;
            }

            int i;

            for (i = 0; i <= 3; i++)
            {
                table[row, col] = i;

                if (i == 0)
                {
                    rows[row].Zeros--;
                    cols[col].Zeros--;

                    if (rows[row].Zeros < 0)
                    {
                        continue;
                    }

                    if (cols[col].Zeros < 0)
                    {
                        continue;
                    }
                }
                else
                {
                    if (i == 1)
                    {
                        rows[row].Zeros++;
                        cols[col].Zeros++;
                    }

                    rows[row].Sum--;
                    cols[col].Sum--;

                    if (rows[row].Sum < 0)
                    {
                        break;
                    }
                    else if (cols[col].Sum < 0)
                    {
                        break;
                    }
                }

                if (col == cols.Length - 1)
                {
                    if (rows[row].Sum != 0 || rows[row].Zeros != 0)
                    {
                        continue;
                    }
                }

                if (row == rows.Length - 1)
                {
                    if (cols[col].Sum != 0 || cols[col].Zeros != 0)
                    {
                        continue;
                    }
                }

                if (!last)
                {
                    solutions += Do(table, rows, cols, nextRow, nextCol);
                }
                else 
                {
                    solutions++;

                    Console.WriteLine("Found solution:");

                    var sums = new int[cols.Length];
                    var zeross = new int[cols.Length];

                    for (int j = 0; j < rows.Length; j++)
                    {
                        int sum = 0;
                        int zeros = 0;

                        for (int k = 0; k < cols.Length; k++)
                        {
                            Console.Write("{0,2} ", table[j, k]);

                            if (table[j, k] == 0)
                            {
                                zeros++;
                                zeross[k]++;
                            }
                            else
                            {
                                sum += table[j, k];
                                sums[k] += table[j, k];
                            }
                        }

                        Console.WriteLine("| Sum {0,2} | Zeros {1}", sum, zeros);

                        Debug.Assert(sum == rows[j].OriginalSum);
                        Debug.Assert(zeros == rows[j].OriginalZeros);
                    }

                    Console.WriteLine("---------------");

                    for (int j = 0; j < cols.Length; j++)
                    {
                        Console.Write("{0,2} ", sums[j]);
                        Debug.Assert(sums[j] == cols[j].OriginalSum);
                    }

                    Console.WriteLine();

                    for (int j = 0; j < cols.Length; j++)
                    {
                        Console.Write("{0,2} ", zeross[j]);
                        Debug.Assert(zeross[j] == cols[j].OriginalZeros);
                    }

                    Console.WriteLine();
                }
            }

            // The for cycle was broken at 0. We have to "readjust" the zeros.
            if (i == 0)
            {
                rows[row].Zeros++;
                cols[col].Zeros++;
            }

            // The for cycle exited "normally". i is too much big because the true last cycle was at 3.
            if (i == 4)
            {
                i = 3;
            }

            // We readjust the sums.
            rows[row].Sum += i;
            cols[col].Sum += i;

            Debug.Assert(oldValueRowSum == rows[row].Sum);
            Debug.Assert(oldValueRowZero == rows[row].Zeros);
            Debug.Assert(oldValueColSum == cols[col].Sum);
            Debug.Assert(oldValueColZero == cols[col].Zeros);

            return solutions;
        }
    }

    public class RowOrCol
    {
        public readonly int OriginalSum;
        public readonly int OriginalZeros;

        public int Sum;
        public int Zeros;

        public RowOrCol(int sum, int zeros)
        {
            this.Sum = this.OriginalSum = sum;
            this.Zeros = this.OriginalZeros = zeros;
        }
    }
}

答案 1 :(得分:1)

它有多快?我只是测试了一个天真的“尝试几乎任何东西”与一些早期的中止但不太可能,并且它非常快(不到一毫秒)。它给出了解决方案:

[[ 0 1 1 1 1 ]
 [ 1 0 1 1 2 ]
 [ 1 0 0 1 2 ]
 [ 2 1 2 2 1 ]
 [ 1 1 0 0 1 ]]

如果这对你来说是一个可以接受的解决方案,我可以发布代码(或者只是讨论它,它非常详细,但基本的想法是微不足道的)

编辑:它也可以简单地扩展到枚举所有解决方案。它在15毫秒内找到了400个,并声称只有这个。这是对的吗?


我做了什么,从0,0开始尝试我可以在那个地方填写的所有值(0但是min(3,rowsum [0])),填充它(从rowsum [y]中减去它如果值为零,则colsum [x]并从rowzero [y]和colzero [x]中减去一个),然后以递归方式为0,1执行此操作; 0,2; 0,3;然后在0.4我有一个特殊情况,我只是填写剩余的rowum,如果它是非负的(否则,中止当前的尝试 - 即在递归树中上升),以及当y = 4时类似的东西。同时,当任何rowum colsum colzero或rowzero变为负数时,我会中止。

当且仅当所有剩余的rowums columnsums colzero和rowzero为零时,当前的板才是解决方案。所以我只测试它,并将其添加到解决方案中,如果它是一个。它不会有任何负面的结构。