Date of Publication :29th March 2018

Abstract: Nonparametric relational topic models provide a successful way to discover the hidden topics from a document network. Most of the theoretical and practical tasks, such as dimensional reduction, document clustering, and link prediction, would benefit from this revealed knowledge. The sampling algorithm scalable to large networks by using new network constrain methods instead of MRFs. Current MRF-based methods do not make the inference efficient enough. Specifically, each document is assigned a Gamma process, although this method provides a solution, it brings additional challenges when mathematically modeling the network structure of typical document network i.e., two spatially closer document stand to have more similar topics. we require the topics are shared the documents through gamma process. In order to resolve these challenges, we use a sub-sampling strategy to assign each and every document a different Gamma process from the global Gamma process, and the sub-sampling probabilities of documents are assigned with a sampling technique instead of Markov Random Field constraint that inherits the document network structure. Through the posterior inference algorithm, we can discover the hidden topics and its number simultaneously. Experimental results on the capabilities of learning the hidden topics and, more importantly, the number of topics.

Reference :