基于在线梯度下降的Mini Batch K-Prototypes算法

MINI BATCH K-PROTOTYPES ALGORITHM USING ONLINE GRADIENT DESCENT

K-Prototypes算法每次迭代都需要花费大量时间来计算所有样本与每个簇中心的相异度以将其划分到各个聚类簇中,这导致K-Prototypes算法在处理大型数据集时运行时间急剧增加。根据在线梯度下降算法对K-Prototypes算法的迭代过程进行优化,减少算法每次迭代所需的计算量从而降低算法的时间复杂度,提升算法运行效率。实验结果表明,提出的基于在线梯度下降的K-Prototypes聚类算法,可以在不影响算法收敛性和有效性的前提下降低算法的时间复杂度,提升算法运行效率。

It requires a lot of time for each iteration of the K-Prototypes algorithm to calculate the dissimilarity degree between all samples and each cluster center when dividing all the samples into different clusters,which results in a sharp increase in the run time for the K-Prototypes algorithm to process large data sets.Therefore,this paper,based on the online gradient descent algorithm,optimizes the iteration process of the K-Prototypes algorithm,so as to reduce the time complexity of the algorithm and improve its running efficiency by reducing the computation required for each iteration of the algorithm.The experimental result shows that the K-Prototypes clustering algorithm proposed in this paper based on online gradient descent can reduce the time complexity of the algorithm and improve its running efficiency without affecting the algorithm convergence and efficiency.