基于高斯混合模型的k均值初始化EM算法的研究

Gaussian mixture model based k-means to initialize the EM algorithm

EM算法是一种非常流行的极大似然估计方法,是一种当观测数据为不完全数据时求解最大似然估计的迭代算法,也是估计有限混合模型参数十分有效的算法.然而,EM算法是一个局部最优算法,常常容易陷入局部最优解,使得它的初始值对算法的结果有着极其重要的影响.因此采用k均值算法来初始化EM算法并将聚类结果同直接用EM算法得到的聚类结果相比较.数值试验表明经过初始化的EM算法的聚类效果要明显好于原始EM算法的效果.

The EM algorithm is a very popular maximum likelihood estimation method, the iterative algorithm for solving the maximum likelihood estimator when the observation data is the incomplete data, but also is very effective algorithm to estimate the finite mixture model parameters. However, EM algorithm is a local optimization algorithm, and often easy to fall into local optimal solution, so the initial value has an extremely important impact on the results of the algorithm. Therefore, we choose the k- means algorithm to initialize the EM algorithm and compare the clustering results with the clustering results obtained through the direct use of the EM algorithm. Numerical experiments show that the clustering effect of the initialization effect of the original EM algorithm. of the EM algorithm is significantly better than the