This is notes which I wrote as I was taking video lecture originated from https://www.youtube.com/watch?v=3S8P3U2ti2A&list=PLbhbGI_ppZISMV4tAWHlytBqNq1-lb8bz&index=70 ================================================================================ ================================================================================ What is the Gibbs sampling used for? ================================================================================ Bayesian network ---> factorization $$$\alpha, \beta$$$: prior knowledge, constant ================================================================================ Perform factorization on following plate notation $$$P(\phi_i;\beta) = P(\phi_i|\beta)P(\beta)$$$ ================================================================================ ================================================================================ Graphical model Factorization using joint ================================================================================ To make above joint equation simpler, remove $$$\theta$$$ and $$$\phi$$$ ================================================================================ To remove $$$\theta$$$ and $$$\phi$$$, you need to use "marginalize out" ================================================================================ About $$$\phi$$$ About $$$\theta$$$ 2 independant integrations ================================================================================ ================================================================================ So far, you performed following ================================================================================ What is the $$$\phi$$$? ================================================================================ ================================================================================ What you have done so far? From (1) and (2), you removed intergrations as much as you could