CSC321: Neural Networks

Lecture 8: The Bayesian way to fit models

The Bayesian framework

A coin tossing example

Some problems with picking the parameters that are most likely to generate the data

Using a distribution over parameter values

Lets do it again: Suppose we get a tail

Lets do it another 98 times

Bayes Theorem

A cheap trick to avoid computing the posterior probabilities of all weight vectors

Why we maximize sums of log probs

A even cheaper trick

Supervised Maximum Likelihood Learning

Supervised Maximum Likelihood Learning