lgalke/mpd-aae-recommender

Applying adversarial autoencoding recommender to Spotify million playlist dataset (RecSysChallenge 2018)

mpd-aae-recommender

Applying adversarial autoencoding recommender to Spotify million playlist dataset

Challenge: RecSys Challenge 2018

Track: Main track

Team name: Unconscious Bias

For more information see Using Adversarial Autoencoders for Multi-Modal Automatic Playlist Continuation

Steps to train a model and apply it to a test set

After cloning the repository it takes very few action to apply our approach. Please make sure to run the code on a machine with GPUs and CUDA support. For the following command line instructions, the current working directory is assumed to be the present git repository.

Step 1: Setup virtual environment and install all dependencies

bash setup.bash

This will create a virtual environment in folder venv and install all the necessary requirements.

Step 2: Activate the virtual environment

source venv/bin/activate

Step 3: Kick-off the experiments (can take a while)

CUDA support is required.

python3 make_submission.py --data-path PATH/TO/MillionPlaylist/data --test-path PATH/TO/MillionPlaylist/test_set.json

Replace the argument for --data-path with the ./data directory of the Spotify Million Playlist Dataset and replace the argument for --test-path with the path to the json file holding test set.

Per default the output will be written to submission.csv, if desired it can be changed by providing -o argument.

Cite

If you use our code in your own work please cite our paper:

@inproceedings{Vagliano:2018,
     author = {Vagliano, Iacopo and Galke, Lukas and Mai, Florian and Scherp, Ansgar},
     title = {Using Adversarial Autoencoders for Multi-Modal Automatic Playlist Continuation},
     booktitle = {Proceedings of the ACM Recommender Systems Challenge 2018},
     series = {RecSys Challenge '18},
     year = {2018},
     isbn = {978-1-4503-6586-4},
     location = {Vancouver, BC, Canada},
     pages = {5:1--5:6},
     articleno = {5},
     numpages = {6},
     url = {http://doi.acm.org/10.1145/3267471.3267476},
     doi = {10.1145/3267471.3267476},
     acmid = {3267476},
     publisher = {ACM},
     address = {New York, NY, USA},
     keywords = {adversarial autoencoders, automatic playlist continuation, multi-modal recommender, music recommender systems, neural networks},
}