Factorization from Conditional Independence for Markov Random Fields | PRML 8.3.2

One reason why undirected graphical models are so useful is that we can read off conditional independence as simple graph separation. In this video we discuss how this implies that the probability distributions we can represent in this way must factorize into a product of factors involving cliques: fully connected subsets of variables. Reference: Bishop, C. M. (2006). Pattern Recognition and Machine Learning. Section 8.3.2: Factorization properties. This video is part of my series reading through Christopher Bishop's PRML. Check out my channel for other chapters in this series.