kmeans聚类 c++实现

/*
num_class: 聚类数
num_data：数据个数
dimension：数据维度（每个数据是多少维的）
data：待聚类数据指针
cluster_center：聚类中心指针
max_error：前后两次误差降低到此值迭代终止
max_iters：最大迭代次数

1、随机初始化聚类中心
2、根据聚类中心计算每个样本属于哪个聚类中心
3、根据聚出来数据重新计算每个类新的聚类中心
重复2，3步骤
*/
void kmeans_cluster(const int num_class, const int num_data, const int dimension, double **data, double **cluster_center, double &max_error, int max_iters = 1000)
{
	//initial cluster center
	//data[num_data][dimension],  cluster_center[num_class][dimension], cannot used as ptr
	//vector<vector<int>> cluster_mark   =>cluster_mark[num_class][num_data_each_class]
	for (int i = 0; i < num_class; i++)
	{
		for (int j = 0; j < dimension; j++)
		{
			cluster_center[i][j] = data[i][j];
		}
	}
	for (int iter = 0; iter < max_iters; iter++)
	{
		double curr_error = 0;
		//calculate class of each data, cluster_mark
		std::vector<std::vector<int>> cluster_mark;
		for (int i = 0; i < num_data; i++)
		{
			double sum[10] = { 0 };//此处并不是想写10，而是dimension，但是数组声明时元素个数不允许用变量
			for (int c = 0; c < num_class; c++)
			{
				for (int d = 0; d < dimension; d++)
				{
					sum[c] = abs(data[i][d] - cluster_center[c][d])+ sum[c];
				}
			}
			cluster_mark[*std::min_element(sum, sum + num_data)].push_back(i);			
		}
		//re-calcute center, cluster_center[num_class][dimension],difference between two iters, max_error
		for (int c = 0; c < num_class; c++)
		{		
			for (int d = 0; d < dimension; d++)
			{
				double mean_d = 0;
				for (int i = 0; i < cluster_mark[c].size(); i++)
				{
					mean_d = mean_d + data[cluster_mark[c][i]][d];
				}
				mean_d = mean_d / cluster_mark[c].size();
				curr_error = curr_error + abs(cluster_center[c][d] - mean_d);
				cluster_center[c][d] = mean_d;
			}
		}

		if (curr_error < max_error)
		{
			break;
		}
	}
}

练习代码，如有错误请指出