EM for Estimating 192#192 Means

Given:

• Instances from 32#32 generated by mixture of 192#192 Gaussian distributions
• Unknown means 193#193 of the 192#192 Gaussians
• Don't know which instance 60#60 was generated by which Gaussian

Determine:

• Maximum likelihood estimates of 193#193

Think of full description of each instance as 194#194, where

• 195#195 is 1 if 60#60 generated by 196#196th Gaussian
• 60#60 observable
• 195#195 unobservable