An alternative to EM for Gaussian mixture models: batch and stochastic Riemannian optimization

Keyphrases

• Gaussian Mixture Model 100%
• Expectation Maximization 100%
• Riemannian Optimization 100%
• Optimization Problem 20%
• Expectation-maximization Algorithm 20%
• Parameter Estimation 20%
• Maximum Likelihood Estimation 20%
• Stochastic Methods 20%
• Positive Definiteness 20%
• Global Analysis 20%
• Stochastic Gradient Algorithm 20%
• Stochastic Gradient 20%
• Positive Definite Matrix 20%
• Riemannian Geometry 20%
• Asymptotic Convergence Analysis 20%
• Nonasymptotic Convergence 20%

Mathematics

• Stochastics 100%
• Gaussian Mixture Model 100%
• Expectation-Maximization Algorithm 25%
• Asymptotics 25%
• Closed Form 25%
• Positive Definite 25%
• Positive Definite Matrix 25%
• Convergence Analysis 25%
• Riemannian Geometry 25%
• Parameter Estimation 25%
• Maximum Likelihood 25%