next up previous contents
Next: Maximum Likelihood Models Up: Probabilistic Models as Data Previous: Probabilistic Models as Data

Models and Data: Some models are better than others

Given a data set and a (generally infinite) set of models, the problem of finding the `best' model in the set to describe the data arises. This is known as point or parameter estimation in the literature. It is not a very well posed problem, as the term `best' is not defined, and there are several possible meanings it might have. In the next sections we look at the best known contenders for `best' models.



Mike Alder
9/19/1997