https://github.com/opringle/collaborative_filtering

A repository to experiment building state of the art recommender systems using apache MXNet
https://github.com/opringle/collaborative_filtering

Last synced: 2 months ago
JSON representation

A repository to experiment building state of the art recommender systems using apache MXNet

• Host: GitHub
• URL: https://github.com/opringle/collaborative_filtering
• Owner: opringle
• Created: 2018-03-21T04:09:33.000Z (almost 7 years ago)
• Default Branch: master
• Last Pushed: 2018-05-17T04:53:25.000Z (over 6 years ago)
• Last Synced: 2024-08-01T22:42:04.281Z (5 months ago)
• Language: Python
• Size: 14.4 MB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• Awesome-MXNet - collaborative_filtering

README

        
# Deep Learning Recommender Systems

A repository to experiment building recommender systems using apache MXNet.

## Repo structure

- `/data` directory used for storing model input data

- `/libs` python modules used by model training script

- `./recysys.py` trains the recommender system 

- `./results` directory used for storing model reports and results

- `./docs`: research notes etc

## Running the code

1. Clone this repo and cd to root

2. Download, unzip and store the ratings data:

    1. `$ mkdir data`

    2. `$ wget 'http://files.grouplens.org/datasets/movielens/ml-1m.zip'` 

    3. `$ unzip -a ml-1m.zip`

    4. `$ mv ./ml-1m/ratings.dat ./data/`

    5. `$ rm -rf ml-1m`

    6. `$ rm ml-1m.zip`

6. Train the model: `$ python recsys.py`

## ToDo

  

- [ ] Use normal data iterators with data in memory

- [ ] Use callback functions to reproduce HR@K plots