Convergence Properties of the K-Means Algorithms

This paper studies the convergence properties of the well known K-Means clustering algorithm. The K-Means algorithm can be described either as a gradient descent algorithmor by slightly extending the mathematics of the EM algorithm to this hard threshold case. We show that the K-Means algorithmactually minimizes the quantization error using the very fast Newton algorithm.

• Type: Conference
• Year: 1994
• Where: NIPS
• Authors: Léon Bottou, Yoshua Bengio
• Keywords: EM Algorithm, K-Means Algorithm, K-means Clustering Algorithm, NIPS 1994, NIPS 2007