10-601 GM3

This a lecture used in the Syllabus for Machine Learning 10-601B in Spring 2016

Slides

• in PPT, in PDF.

Readings

• See first lecture on GM
• For EM: Mitchell 6.2 or Murphy 11.4.1, 11.4.2, 11.4.4

To remember

• The EM algorithm
  • E-step (expectation step)
  • M-step (maximization step)
• How to use EM to learn DGMs with hidden variables
• How to use EM to learn a mixture of Gaussians

• Connections:
  • naive Bayes as a DGM
  • semi-supervised naive Bayes as a DGM with hidden variables
  • mixture of Gaussians as an a DGM
  • mixture of Gaussians vs k-means