Applying adversarial autoencoding recommender to Spotify million playlist dataset (RecSysChallenge 2018)
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
aae-recommender @ c6a3d0d
.gitignore
.gitmodules
LICENSE
README.md
create_dev_set.py
evaluate_dev.py
make_submission.py
mpd.py
mpd_metrics.py
mpd_stats.py
setup.bash

README.md

mpd-aae-recommender

Applying adversarial autoencoding recommender to Spotify million playlist dataset

Challenge: RecSys Challenge 2018

Track: Main track

Team name: Unconscious Bias

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.