基于小世界模型的高维数据查询算法

摘要

Based on small-world model, express the high-dimensional feature vector as the network nodes, and then design the high-dimensional index generation algorithm based on K-Means technology and the random approximate neighbor query algorithm. With the appropriate chosen the number of neighbor nodes, the maximum length of query paths and the maximum iterations, the proposed algorithm can meet various query with different precision demands. Experiment demon-strates the algorithm can achieve effective index performance with mass high-dimensional data vectors.%提出了一种从海量高维数据中进行高效查询的算法,该算法基于小世界网络模型,并采用网络节点表示高维数据的特征向量.算法主要包含两个部分,基于K-Means的索引生成算法和随机逼近查询算法,两个算法均给出了具体的操作步骤.算法经大量实验仿真,得出通过合理设置小世界网络节点的近邻节点数量以及最大查询路径和最大迭代次数等参数,算法可以满足不同精度的用户查询请求.实验结果表明,实现的算法在高维度海量数据查询中具有良好的检索效果.