Lecture 23 Dirichlet Processes


— Handwritten subtitles — Generated subtitles

Start Reading:

hey good morning every everyone so how you guys doing still staying safe and unhappy so I hope well getting more used to this remote asian mode and feeling comfortable so today we are going to continue to cover complex distributions if you recall last two lectures we covered distribution over functions which is a Gaussian process and then we discussed a distribution over sets which is the determinant of point process and today and the next we are going to discuss distributions over over distributions in fact and also distributions over distributions of features so these are very used for tools to further enhance the graphic model to do really fancy stuff and sometimes in a very very flexible computationally and also modeling wise fashion and also these tools are really composable and integrate always existing methods such as deep neural networks and other tools so we will do them as a building block for that other task alright so let begin by you know a classical problem to motivate this new technique so we all know the problem of clustering so here I show you the data points spread in two dimensional space and I want you to