EM algorithm: |
---|
Step1: |
The number of categories K is preset as 3, then set the initial values of \(\theta\) for each component K and calculate the log-likelihood value in Eq. (7) |
Step2: E step |
Based on current values of \(\theta\), the value of z for each sample is estimated |
Step3:M step |
The values of z in Eq. (7) are updated, and the log-likelihood value is maximized to get a new set of \(\theta\) values |
Step4: |
Return to Step 2 until convergence |