Gaussian Mixture Model: Expectation-Maximization Clustering (2D)

This simulator fits a Gaussian Mixture Model to a 2D point cloud with real Expectation-Maximization, drawn directly on a pannable, zoomable 2D canvas. Each of the K components is rendered as a colored ellipse — the 2σ contour of its own mean and covariance — and every data point's color is a genuine blend of its soft "responsibilities" across all components (or, in hard-assign mode, its single most likely component). Step through E+M iterations one at a time or let it play automatically, regenerate the underlying blobs, reinitialize the components from scratch, tune how many points and clusters are in play, and watch the log-likelihood climb on a live strip chart as the fit converges.