I'm going over the paper by Hoffman et. al. titled "Stochastic Variational Inference" (SVI) where the authors present an algorithm for Hierarchical Dirichlet Process (HDP) topic models. I'm comparing this with a paper by Chong Wang et. al. titled "Online Variational Inference for the Hierarchical Dirichlet Process" where the authors present an online HDP algorithm (also implemented in gensim). And I'm wondering between the differences of these two algorithms. In what ways SVI HDP is different from online HDP?
There is not a difference; the online method in Chong et al. is an instance of the method later generalized in Hoffman et al., a fact that is tucked into a footnote in the SVI paper:
This section is organized as follows. We first give some background on the Dirichlet process and its definition via Sethuraman’s stick breaking construction, which is a distribution on the infinite simplex. We then show how to use this construction to form the HDP topic model and how to use stochastic variational inference to approximate the posterior.$^7$
- This algorithm first appeared in Wang et al. (2011). Here we place it in the more general context of Section 2 and relate it to stochastic inference for LDA.