variational mixture of Gaussians
(2.7 hours to learn)
Summary
Variational Bayes EM can be applied to fitting a mixture of Gaussians model. Unlike standard mixture of Gaussians fit with EM, the variational algorithm automatically controls the model complexity and yields a lower bound on the marginal likelihood.
Context
This concept has the prerequisites:
- variational Bayes
- mixture of Gaussians models
- Bayesian parameter estimation: multivariate Gaussians (The posterior over parameters for a multivariate Gaussian is part of the update rule for this model.)
- Bayesian parameter estimation: multinomial distribution (The posterior over the mixture probabilities is part of the update rule for this model.)
Core resources (read/watch one of the following)
-Paid-
→ Pattern Recognition and Machine Learning
A textbook for a graduate machine learning course, with a focus on Bayesian methods.
Location:
Section 10.2, pages 474-486
Supplemental resources (the following are optional, but you may find them useful)
-Paid-
→ Machine Learning: a Probabilistic Perspective
A very comprehensive graudate-level machine learning textbook.
Location:
Section 21.6, pages 749-756
See also
-No Additional Notes-