https://github.com/lyst/lightfm-paper
https://github.com/lyst/lightfm-paper
Last synced: over 1 year ago
JSON representation
- Host: GitHub
- URL: https://github.com/lyst/lightfm-paper
- Owner: lyst
- Created: 2015-07-30T08:34:08.000Z (almost 11 years ago)
- Default Branch: master
- Last Pushed: 2015-07-31T14:16:07.000Z (almost 11 years ago)
- Last Synced: 2025-01-10T22:43:29.307Z (over 1 year ago)
- Language: TeX
- Size: 648 KB
- Stars: 45
- Watchers: 6
- Forks: 15
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Metadata Embeddings for User and Item Cold-start Recommendations
Pre-print available on [arXiv](http://arxiv.org/abs/1507.08439).
## Structure
1. The `paper` directory holds the LaTeX source of the paper.
2. The `experiments` directory holds the code used for experiments.
## Reproducing results
To reproduce the results from the paper, do the following.
1. Download and install the LightFM package (`pip install lightfm`).
2. Download experiment data by running `cd experiments/stackoverflow/ && make` and `cd experiments/movielens/ && make`.
3. To run the primary experiment, run `ipython -- runner.py --table`. This will print LaTeX variable definitions that, when pasted into the paper source file, populate the main results table.
4. To generate the latent dimension sensitivity plots, run `ipython -- runner.py --dim --plot`. This will take some time.
It is possible to run invididual experiments. Examples:
- `ipython -- experiments/stackexchange/model.py --dim 50 --lsi --tags --ids --split 0.2` will run the CrossValidated experiment with 50-dimensional
latent space, using the LSI-LR model with both post tags and post ids.
- `ipython -- experiments/movielens/model.py --ids --cold --split 0.2` will run the MF model on the MovieLens dataset in the cold-start scenario.