An improved AD-LDA topic model based on weighted Gibbs sampling
Hongchen Guo, Qiliang Liang, Zhiqiang Li · 2016
Topic models such as LDA have been widely used to capture latent topics in textual collections. Due to the large scale of data, researchers begin to pay more attention to speed up the efficiency of LDA Gibbs sampling by parallel inference algorithms or distributed computing. In this paper, we improve the traditional Approximate Distributed LDA (AD-LDA) algorithm by introducing weighted factor during iteration of Gibbs sampling, and propose a novel model called Weighted AD-LDA (WAD-LDA). Experiments are conducted over real-world dataset and the results show that our method can promote the computing efficiency significantly and can achieve meaningful topics as AD-LDA simultaneously.