Assessment info The assessment will not cover anything from now on; will cover from lecture 14 (?) back. No lecture Tuesday Week 10  More/recap on Gibbs sampling Gibbs sampling with already instantiated variables doesn’t always work. For example: take a … Continue reading →

What is this? Another form of sampling, used in bayesian networks etc. In previous lecture we talked about forward, rejection, and importance sampling. So why do we need Gibbs sampling too? Limitations of importance sampling If the evidence is improbable … Continue reading →

We covered some of this last lecture; how to get random numbers in Python. Why sample? Some problems can be solved, but would take ages. Instead we can write a sampler to simulate the problem x number of times and … Continue reading →

Given a join tree, how would we computer the probability of just ‘A’ for example. With variable elimination we’d have to sum it all out. With a join tree however, we can look at the structure of the data. There … Continue reading →

Previously Triangulating a graph. Is it me or does that just mean removing squares in the graph (yes, no cycles of 4 or more)? You do it by removing vertexes, and you have to add fill in lines. Where is … Continue reading →

(Sometimes called ‘The Sum Product Algorithm’) The basic ‘inference problem’ in graphical models is – given a joint distribution, you want to compute the marginal distribution for a given variable. One option – multiply out all factors and marginalise on … Continue reading →

Summary of how far we got Have been talking about factors and how one can marginalise factors (removing variables) but also do multiplication and division of them. Probably should look up why you do multiplication and division of factors? I … Continue reading →