Gaussian Mixture Model

Lecture Note

GMM-EM class note

Gaussian Mixture Model

Click here for PPT slides: TAMU

Click here for PPT slides: VTech

Gaussian Distribution

The density can be estimated by multiples of Gaussian Kernels

$\pi$is prior, not likelihood.

The mixing coefficients are themselves probabilities and must meet this condition: sum(pi)=1

How to find the optimal parameter $\Theta$?

Lets use the Maximum Likelihood Estimation to find the optimal parameter. This can be solved by E-M algorithm/

Example: Object segmentation

see PPT slides: VTech

E-M algorithm for Gaussian Mixture Kernel

Read here for detail

Reference

The Hundred-Page Machine Learning Book http://themlbook.com/wiki/doku.php

머신러닝/패턴인식, 오일석

A Comprehensive Overview of Gaussian Splatting | Towards Data ScienceTowards Data Science