VAE-for-Collaborative-Filtering-Pytorch

Implementation of 'Variational Autoencoders for Collaborative Filtering' paper [https://arxiv.org/abs/1802.05814] in Pytorch

This repository include:

1. Data Processing
2. Denoising AutoEncoder Model in the paper
3. Validation for the model
4. Model Ablation Studies: a. Explicit Use of l2 norm for model weights b. Effect of dropout

Dependency: Python3

Simply run the git_main.ipynb

Validation Study for the Models:

Validation for new model currently developing

Will upload the code when the results are comparable

Final: Model with explicit l2 loss no_l2 : Model without explicit l2 loss no_drop: Model without dropout. It overfits as expected.

To DO:

1. Clean up the code
2. Add VAE model
3. Add test matrix in the code

Note: Data preparation is mostly based on original implementation source with some changes.