I was going through lda2vec and was confused on some of the concepts.It is a combination of LDA and word2vec.Word2vec is used to learn dense word vectors and LDA is used to learn the probability distribution over topics for each document.As a combination,the vector representation for each topic should also be dense right? I was kind of confused on the requirement of dense/sparse vector in word2vec and related algorithms. In LDA,is it required that the dirichlet priors on the pre-document topic/word distributions to be non-sparse? If so, what would happen if the assumption is not satisfied? Any help will be appreciated!