I would like to compare between different bayesian topic models such as LDA with seeding, hierarchical, hidden markov topic modeling, gaussian hmm on sentence embedding etc. I look for a platform for implementing these varients. Is there past experience on topic modeling using Edward in the community? An LDA example would be useful with Gibbs sampling.
Implementing mixed membership models
I added an example in the Github repo,
examples/deep_exponential_family.py, which is one of the state of the art for topic modeling. You might try to build from that code.
Alternatively if you’re into classics, you can build LDA + collapsed Gibbs by extending