Timezone: »
In this paper, we study convergence properties of the gradient variant of Expectation-Maximization algorithm~\cite{lange1995gradient} for Gaussian Mixture Models for arbitrary number of clusters and mixing coefficients. We derive the convergence rate depending on the mixing coefficients, minimum and maximum pairwise distances between the true centers, dimensionality and number of components; and obtain a near-optimal local contraction radius. While there have been some recent notable works that derive local convergence rates for EM in the two symmetric mixture of Gaussians, in the more general case, the derivations need structurally different and non-trivial arguments. We use recent tools from learning theory and empirical processes to achieve our theoretical results.
Author Information
Bowei Yan (Jump Trading)
Mingzhang Yin (University of Texas at Austin)
Purnamrita Sarkar (UT Austin)
More from the Same Authors
-
2018 Poster: Overlapping Clustering Models, and One (class) SVM to Bind Them All »
Xueyu Mao · Purnamrita Sarkar · Deepayan Chakrabarti -
2018 Spotlight: Overlapping Clustering Models, and One (class) SVM to Bind Them All »
Xueyu Mao · Purnamrita Sarkar · Deepayan Chakrabarti -
2018 Poster: Mean Field for the Stochastic Blockmodel: Optimization Landscape and Convergence Issues »
Soumendu Sundar Mukherjee · Purnamrita Sarkar · Y. X. Rachel Wang · Bowei Yan -
2017 Poster: On clustering network-valued data »
Soumendu Sundar Mukherjee · Purnamrita Sarkar · Lizhen Lin -
2016 Poster: On Robustness of Kernel Clustering »
Bowei Yan · Purnamrita Sarkar -
2015 Poster: The Consistency of Common Neighbors for Link Prediction in Stochastic Blockmodels »
Purnamrita Sarkar · Deepayan Chakrabarti · peter j bickel
