Kmeans++算是DONet实现

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Drawing;
using System.Drawing.Imaging;
namespace ConsoleApplication1
{
    class Program
    {
        static void Main(string[] args)
        {
            Random random = new Random(Math.Abs(unchecked((int)DateTime.Now.Ticks)));
            List<Coordinate> rawData = new List<Coordinate>();
            for (int i = 0; i < 500; i++)
            {
                rawData.Add(new Coordinate { X = random.Next(0, 500), Y = random.Next(0, 500) });
            }
            
            KmeansPlus plus = new KmeansPlus();
            plus.Data = rawData;
            plus.K = 10;
            plus.ShortDistance = 20;
            plus.Start();
            DrawPoint(rawData, plus.Means, plus.Clustering);
            Console.ReadLine();
        }
        
        /// <summary>
        /// 绘制图片显示出来
        /// </summary>
        /// <param name="vector"></param>
        /// <param name="mean"></param>
        /// <param name="clustering"></param>
        static void DrawPoint(List<Coordinate> vector, List<Coordinate> mean, int[] clustering)
        {
            Bitmap bit = new Bitmap(600, 600);
            Graphics g = Graphics.FromImage(bit);
            g.Clear(Color.White);
            SolidBrush Black = new SolidBrush(Color.Black);
            SolidBrush Red = new SolidBrush(Color.Red);
            List<Color> ColorList = new List<Color>();
            ColorList.Add(Color.MediumOrchid);
            ColorList.Add(Color.DeepPink);
            ColorList.Add(Color.Blue);
            ColorList.Add(Color.Brown);
            ColorList.Add(Color.Coral);
            ColorList.Add(Color.CornflowerBlue);
            ColorList.Add(Color.DarkCyan);
            ColorList.Add(Color.DarkGreen);
            ColorList.Add(Color.DarkMagenta);
            ColorList.Add(Color.DarkRed);
            ColorList.Add(Color.DodgerBlue);
            Font f = new Font("宋体", 10);
            for (int i = 0; i < vector.Count; i++)
            {
                Coordinate p = vector[i];

                g.FillEllipse(Black, Convert.ToInt32(p.X), Convert.ToInt32(p.Y), 5, 5);



            }

            for (int i = 0; i < vector.Count; i++)
            {
                Coordinate p = vector[i];
                int color = clustering[i];
                g.DrawString(color.ToString(), f, new SolidBrush(ColorList[color]), Convert.ToInt32(p.X), Convert.ToInt32(p.Y));



            }
            int j = 0;
            foreach (Coordinate p in mean)
            {

                g.FillEllipse(new SolidBrush(ColorList[j]), Convert.ToInt32(p.X), Convert.ToInt32(p.Y), 8, 8);
                j++;
            }
            bit.Save(@"E:\Debug\" + DateTime.Now.Millisecond + ".png", ImageFormat.Png);
        }
    }

}

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;

namespace ConsoleApplication1
{
    public class Kmeans
    {
        public int[] Cluster(List<Coordinate> rawData, List<Coordinate> means, ref List<Coordinate> endmeans)
        {


            List<Coordinate> data = rawData;

            bool changed = true; 

            int[] clustering = null; 

            int maxCount = data.Count * 20; // sanity check
            int ct = 0;
            while (changed == true && ct < maxCount)
            {
                ++ct;
                int[] _clustering = new int[data.Count];
                List<Coordinate> _means = new List<Coordinate>();
                changed = UpdateMeans(data, means, clustering, ref _clustering, ref _means);
                means = _means;
                
                clustering = _clustering;
            }
            endmeans = means;

            return clustering;
        }

        /// <summary>
        /// 计算核心点和聚类点。
        /// </summary>
        /// <param name="data">原始数据</param>
        /// <param name="means">上一次的中心点</param>
        /// <param name="clustering">上一次的聚类点</param>
        /// <param name="newclustering">返回新的中心点</param>
        /// <param name="newmeans">返回新的聚类点</param>
        /// <returns>是否发生更新。</returns>
        private static bool UpdateMeans(List<Coordinate> data, List<Coordinate> means, int[] clustering, ref int[] newclustering, ref List<Coordinate> newmeans)
        {
            newclustering = new int[data.Count];
            int[] clusterCounts = new int[means.Count];
            for (int i = 0; i < data.Count; i++)
            {
                int _clusting = MinIndex(data[i], means);
                newclustering[i] = _clusting;
                clusterCounts[_clusting]++;
            }
            List<Coordinate> _means = new List<Coordinate>();
            for (int i = 0; i < means.Count; i++)
            {
                Coordinate p = new Coordinate { X = 0, Y = 0 };
                _means.Add(p);

            }
            for (int i = 0; i < data.Count; ++i)
            {
                int cluster = newclustering[i];
                _means[cluster].X += data[i].X; // accumulate sum
                _means[cluster].Y += data[i].Y; // accumulate sum
            }
            newmeans = new List<Coordinate>();
            for (int k = 0; k < _means.Count; ++k)
            {

                double x = _means[k].X / clusterCounts[k]; // danger of div by 0
                double y = _means[k].Y / clusterCounts[k]; // danger of div by 0
                Coordinate p = new Coordinate { X = x, Y = y };
                newmeans.Add(p);
            }
            if (clustering == null)
            {
                return true;
            }
            else
            {
                for (int i = 0; i < newclustering.Length; i++)
                {
                    if (newclustering[i] != clustering[i])
                    {
                        return true;
                    }
                }
            }
            return false;
        }
       
        /// <summary>
        /// 计算点到核心点距离获取最小距离点索引；
        /// </summary>
        /// <param name="p"></param>
        /// <param name="means"></param>
        /// <returns></returns>
        private static int MinIndex(Coordinate p, List<Coordinate> means)
        {
            double[] distances = new double[means.Count];
            for (int i = 0; i < means.Count; i++)
            {
                distances[i] = Distance(p, means[i]);
            }
            int indexOfMin = 0;
            double smallDist = distances[0];
            for (int k = 0; k < distances.Length; ++k)
            {
                if (distances[k] < smallDist)
                {
                    smallDist = distances[k];
                    indexOfMin = k;
                }
            }
            return indexOfMin;
        }
        /// <summary>
        /// 距离计算
        /// </summary>
        /// <param name="tuple"></param>
        /// <param name="mean"></param>
        /// <returns></returns>
        public  static double Distance(Coordinate tuple, Coordinate meas)
        {
            double sumSquaredDiffs = 0.0;
            sumSquaredDiffs += Math.Pow((tuple.X - meas.X), 2);
            sumSquaredDiffs += Math.Pow((tuple.Y - meas.Y), 2);
            return Math.Sqrt(sumSquaredDiffs);
        }
    }
}

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;

namespace ConsoleApplication1
{
    public class KmeansPlus
    {
        private int _k = 10;
        private int _shortdistance = 50;
        
        /// <summary>
        /// 初始设置值
        /// </summary>
        public int K { get { return this._k; } set { this._k = value; } }
        /// <summary>
        /// 最短距离合并
        /// </summary>
        public int ShortDistance { get { return this._shortdistance; } set { this._shortdistance = value; } }
        /// <summary>
        /// 初始数据
        /// </summary>
        public List<Coordinate> Data { get; set; }
        /// <summary>
        /// 返回中心点信息
        /// </summary>
        public List<Coordinate> Means { get; set; }
        /// <summary>
        /// 数据分组信息
        /// </summary>
        public int[] Clustering { get; set; }
        public void Start()
        {
            Random random = new Random(Math.Abs(unchecked((int)DateTime.Now.Ticks)));


            List<Coordinate> means = new List<Coordinate>();

            int maxlength = Data.Count - 1;
            int measpoint = random.Next(0, maxlength);
            means.Add(Data[measpoint]);


            List<Coordinate> result = new List<Coordinate>();
            for (int i = 0; i < Data.Count; i++)
            {
                result.Add(Data[i]);
            }
            result.Remove(Data[measpoint]);
            for (int i = 0; i < K; i++)
            {
                List<Coordinate> nextresult = new List<Coordinate>();
                means = UpdateCompass(result, means, ref nextresult);
                result = nextresult;
            }
            
            means = MergeMeans(means, ShortDistance);

            List<Coordinate> Lmeans = new List<Coordinate>();
            int[] clustering = new Kmeans().Cluster(Data, means, ref Lmeans);


            Clustering = clustering;
            Means = Lmeans;
        }
        /// <summary>
        /// 罗盘法随机核心点
        /// </summary>
        /// <param name="data">点</param>
        /// <param name="meas">中心点</param>
        /// <param name="result">去除核心点的数组，下一次使用</param>
        /// <returns></returns>
        private static List<Coordinate> UpdateCompass(List<Coordinate> data, List<Coordinate> meas, ref List<Coordinate> result)
        {
            double[] distance = new double[data.Count];
            result = new List<Coordinate>();
            double sumlength = 0;
            for (int j = 0; j < data.Count; j++)
            {
                double[] _distance = new double[meas.Count];
                for (int i = 0; i < meas.Count; i++)
                {
                    _distance[i] = Kmeans.Distance(data[j], meas[i]);
                }
                double min = _distance.Min();
                distance[j] = min;
                sumlength += min;
            }
            Random random = new Random(Math.Abs(unchecked((int)DateTime.Now.Ticks)));
            double measpoint = random.Next(0, Convert.ToInt32(sumlength));

            int _postion = 0;
            for (int j = 0; j < distance.Length; j++)
            {
                if ((measpoint - distance[j]) <= 0)
                {
                    _postion = j;

                }
                measpoint -= distance[j];
            }
            meas.Add(data[_postion]);

            data.Remove(data[_postion]);
            result = data;

            return meas;
        }

        /// <summary>
        /// 合并中心
        /// </summary>
        /// <param name="meas">中心点</param>
        /// <param name="len">合并最小距离</param>
        /// <returns></returns>
        private static List<Coordinate> MergeMeans(List<Coordinate> means, int len)
        {

            for (int i = 0; i < means.Count - 1; i++)
            {
                for (int j = 1; j < means.Count; j++)
                {
                    if (i == j)
                    {
                        continue;
                    }
                    double l = Kmeans.Distance(means[i], means[j]);
                    if (l < len)
                    {
                        means.Remove(means[j]);
                        j--;
                    }
                }
            }

            return means;
        }
        //private static double Distance(Coordinate tuple, Coordinate meas)
        //{
        //    double sumSquaredDiffs = 0.0;
        //    sumSquaredDiffs += Math.Pow((tuple.X - meas.X), 2);
        //    sumSquaredDiffs += Math.Pow((tuple.Y - meas.Y), 2);
        //    return Math.Sqrt(sumSquaredDiffs);
        //}
    }
}

 

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;

namespace ConsoleApplication1
{
     [Serializable]
    public class Coordinate
    {
        public double X { get; set; }
        public double Y { get; set; }
    }
}