- basics:
-
maximum likelihood is a popular optimization criterion
-
assume an underlying probability model
-
estimate parameters of the model by choosing the distribution
which has the highest likelihood of having produced the data
- pros:
-
efficient in the case where estimation is completely data
driven
-
efficient iterative estimation algorithms available (EM)
- cons:
-
how do we choose the form of the models?
-
Gaussians typically used since mean and variance
easily estimated -- what about higher order moments?
-
cannot explicitly incorporate additional knowledge
about the data set