Running the code

1. The main result of this paper can be reproduced via running the scripts/main_vae.py file:

   cd scripts && python main_vae.py

2. Download the dataset you want to run with

   python setup_data.py --dataset=DATASET_OF_CHOICE

Default is ml-20m. Other options are amazon, which is my original furniture review data from amazon review data.

Main Scripts

All of the scripts used to generate data are located in the ./scripts directory.

• main_vae.py is MultiVAE model with and without the critic, on all datasets.
• main_wmf.py is used to generate results for the Weighted Matrix Factorization model on all 3 datasets.
• main_dae.py is used to generate results for MutliDAE on ml-20m with and without the Critic
• ncf_training.py is used to generate results on the datasets we compare to Neural Collaborative Filtering.
• gather_data.py is used to gather statistics from a fully trained model,.
• setup_data.py is used to download, extract, and process the data.