Hacker News new | ask | show | jobs
by platz 3852 days ago
Koller's PGM was fun but very hard to generalize outside of the problems presented in the class. Opportunities to implement loopy belief propagation just don't seem to come up as much ad I'd like.
1 comments

That's because most were toy examples dealing with discrete distributions in a few dimensions. Granted, these are mathematically easier to deal with, but not representative of real-world scenarios.