The European Conference on Machine Learning and Principles and Practice of Knowledge Discovery (ECML-PKDD 2016) will be held at Riva del Garda, Italy, during September 19–23, 2016. The following oral presentation will be delivered.
In the field of machine learning, a topic analysis and cluster analysis have been widely explored for long years, however they are developed independently. In our paper, we propose a cHDP, a non-parametric Bayesian model for simultaneous optimization in both clustering and topic modeling. Compared to the state of the art models in the “simultaneous” approach, our model outperforms them in prediction accuracy and computational cost. We verified our model with real corpus data and also applied our model to urban dynamics (daily people flow in a city) analysis with gps-enabled smartphone location data of million people.
— presentation information —
Conference Track 2016-09-21 10:50 – 12:50
Coupled Hierarchical Dirichlet Process Mixtures for Simultaneous Clustering and Topic Modeling
Masamichi Shimosaka*, Takeshi Tsukiji**, Shoji Tominaga**, Kota Tsubouchi***
(*Tokyo Institute of Technology, **The University of Tokyo, ***Yahoo! JAPAN Research, Japan)
We also update the project page.