SOCR_EM_MixtureModelChartDemo

This experiments demonstrates the Expectation Maximization (EM) algorithm. In this setting the EM is applied as a tool for classisfication.

  1. Select random points in the 2D plane (manually by clicking the mouse in the field of view or by clicking the RandomPntsbutton.
  2. Then select the number of cluster that you want to identify
  3. Click InitKernels to get a different starting condition (EM algorithm is VERY sensitive to the starting conditions!)
  4. Select Normal/Fast/Slow spead of the algorithm (for demo purposes choose Slow)
  5. Choose Gaussian or Linear fit for your mixture model
  6. Click EM Run to start the algorithm. Observe the evolution of the process (convergence is guaranteed!)
  7. Finally, use EM Stop or EM 1 Step to terminate the or take one step at a time

You can Segmen+ "This experiments demonstrates the Expectation Maximization (EM) algorithm. In this setting the EM is applied as a tool for classisfication.

  1. Select random points in the 2D plane (manually by clicking the mouse in the field of view or by clicking the RandomPnts button.
  2. Then select the number of cluster that you want to identify
  3. Click nitKernels to get a different starting condition (EM algorithm is VERY sensitive to the starting conditions!)
  4. Select Normal/Fast/Slow spead of the algorithm (for demo purposes choose Slow)
  5. Choose Gaussian or Linear fit for your mixture model
  6. Click EM Run to start the algorithm. Observe the evolution of the process (convergence is guaranteed!)
  7. Finally, use EM Stopor EM 1 Step to terminate the or take one step at a time You can Segment the initial points based on your Linear/Gaussian fit by pressing Segmentthe initial points based on your Linear/Gaussian fit by pressing ";