EM Algorithm — Gaussian Mixture Model
Iterative E-step (soft assignments) + M-step (parameter update)
Controls
Components K:
3
Data points:
120
New Dataset
E-step + M-step
Run to convergence
Reset parameters
Iteration:
0
Log-likelihood:
—
E-step:
Compute soft assignments γ(z_nk) — how much each point "belongs" to each Gaussian.
M-step:
Update μₖ, σₖ, πₖ to maximize expected log-likelihood.
Converges to local maximum of p(X|θ).