Week 11 Human Ethics
We said earlier that LDA assumes that there is some abstract generation process going on behind the scenes, in which hidden (latent) topics generate words that make up the documents in the collection that you observe. The algorithm's main goal is then to reverse-engineer this generation process and to identify the components of this hidden structure-the topics, their distribution, and the distribution of words in each topic. Because of this background assumption that the data is generated by certain parameters, LDA is called a generative model. LDA is not the only algorithm that assumes that a generative process is behind the data observed and that tries to estimate the parameters responsible for such generation-in fact, there is a whole family of machine-learning algorithms called generative models
Larry Shi