I chose to write my Bachelor thesis on LDA with an econometric approach. I have no idea what LDA was / is. Thank you so much for this video I highly enjoyed it and learned a lot. Thanks!
Because probability distributions sum to 1, you can use a triangle for a 3D distribution. You're right that if you add an additional dimension, you'd then need a tetrahedron: mathworld.wolfram.com/Simplex.html
Hi.. thanks for this nice video.. really helpful.. i have a question... after identification of topics/clusters of all docs, if new doc came do we need to run whole process again.. or there is any other way to directly map new doc with pre defined clusters/Topics..
Sorry for the slow reply! If you have a set of topics you like, you can do inference on the new documents. (I.e., do Gibbs sampling, don't let the topics change.) However, if you want the topics to change, then that's more complicated.