Code for the article: 
O. Gouvert, T. Oberlin, C. Févotte "Recommendation from Raw Data with Adaptive Compound Poisson Factorization"

Implemented in Python 2.7.

The repository contains:

- model: the algorithms for the models described in the article, namely discrete compound Poisson factorization (dcPF), for different choices of element distributions: 
    * logarithmic (dcpf_Log)
    * zero-truncated Poisson (dcpf_ZTP)
    * shifted geometric (dcpf_Geo) 
    * shifted negative binomial (dcpf_sNB)

- dataset: the subset of the Taste Profile dataset used in the article

- script_dcpf_TPS.py is the script used for the experiments of the article

- demo_dcpf.py is a small example of application of dcPF on synthetic data

- In "function": some functions to calculate scores, borrowed from Dawen Liang