The Matlab Code for the ICML 2015 paper "Scalable Deep Poisson Factor Analysis for Topic Modeling"
Zhe Gan (zhe.gan@duke.edu), Changyou Chen (cchangyou@gmail.com), Ricardo Henao (ricardo.henao@duke.edu), David Carlson (david.carlson@duke.edu), Duke University, ECE department, 5.4.2015.
This code implement the deep Poisson factor analysis (DPFA), from the 2015 ICML paper with the same title. This source code is made publicly available for reproducibility purposes, it is not optimized for speed, minimally documented but fully functional.
I didn't try to integrate the codes, so each dataset has a separate file...
Please note that this code should be used at your own risk. There is no implied guarantee that it will not do anything stupid. Permission is granted to use and modify the code.
Please cite our ICML paper in your publications if it helps your research:
@inproceedings{DPFA_ICML2015,
Author = {Z. Gan, C. Chen, R. Henao, D. Carlson, and L. Carin},
Title = {Scalable Deep Poisson Factor Analysis for Topic Modeling},
booktitle={ICML},
Year = {2015}
}
The 20news and RCV2 dataset we used can be downloaded from https://drive.google.com/drive/u/0/folders/0B1HR6m3IZSO_WkdFWTdVU1JjMWM
The WIKI dataset are downloaded using scripts provided in Hoffman et al. (2010).