Practical Implementation for -means EM

• is just the probability that instance was generated by the th Normal distribution.
• In the first step,
• In the second step, , similar to the formula for a single Normal distribution. The new formula just has each instance weighted by the expectation that it was generated by the th Normal distribution.
• The algorithm will converge to a local maximum likelihood hypothesis for
• The General EM Algorithm without simplifications is in the book.