Metadata Embeddings for User and Item Cold-start Recommendations

Maciej Kula
7/30/2015
cs.IR

Official code: lyst/lightfm (5,116 stars, Apache-2.0, last commit 2024-07-24). 1 more implementation below.

Abstract

I present a hybrid matrix factorisation model representing users and items as linear combinations of their content features' latent factors. The model outperforms both collaborative and content-based models in cold-start or sparse interaction data scenarios (using both user and item metadata), and performs at least as well as a pure collaborative matrix factorisation model where interaction data is abundant. Additionally, feature embeddings produced by the model encode semantic information in a way reminiscent of word embedding approaches, making them useful for a range of related tasks such as tag recommendations.

View on arXivView PDF

Code Implementations(2)

lyst/lightfmOfficial100%
Apache-2.0

A Python implementation of LightFM, a hybrid recommendation algorithm.

5,116724Python, MakefileJul 30, 20152 years agoApache-2.0
learning-to-rankmachine-learningmatrix-factorizationpythonrecommender+1 more
lyst/lightfm-paperOfficial100%
4515TeX, PythonJul 30, 201511 years ago

Similar Papers

Cross-Attribute Matrix Factorization Model with Shared User Embedding

82% match2023

Scalable Recommendation with Poisson Factorization

80% match1 repo552013

Bayesian Hybrid Matrix Factorisation for Data Integration

79% match1 repo132017

Cross-Domain Latent Factors Sharing via Implicit Matrix Factorization

78% match1 repo22024

GLocal-K: Global and Local Kernels for Recommender Systems

78% match1 repo632021

UIPC-MF: User-Item Prototype Connection Matrix Factorization for Explainable Collaborative Filtering

77% match2023

Cite this paper

@article{kula2015metadata,
  title  = {Metadata Embeddings for User and Item Cold-start Recommendations},
  author = {Maciej Kula},
  year   = {2015},
  eprint = {1507.08439},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1507.08439}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1507.08439.svg)](https://codeofpaper.com/paper/1507.08439)

Discussion