EM Algorithm

Abstract

This chapter is concerned with the EM algorithm which is a frame of clustering data items, rather than a specific algorithm. We will survey some probabilistic distributions as the basis for understanding the EM algorithm. Assuming that each cluster has its own normal distribution, we describe in detail the EM algorithm, which is extended from the k means algorithm. We mention the variations of EM algorithm and the semi-supervised learning algorithm with its combination with the Naive Bayes. From Chaps. 9–11, we cover the kinds of clustering algorithms, and in Chap. 12, we do the clustering index as a clustering evaluation metric.

In Sect. 11.1, we provide the general concept of EM algorithm, and in Sect. 11.2, we mention the various distributions which are assumed as the cluster distributions. In Sect. 11.3, we describe the specific EM algorithm where each cluster is assumed as the Gaussian distribution, entirely. In Sect. 11.4, we mention variants of EM algorithm and in Sect. 11.5, we make the summarization on this chapter and the further discussions. This chapter is intended to describe the EM algorithm which is advanced from the k means algorithm, entirely.