On the instability of embeddings for recommender systems: the case of matrix factorization.
Giovanni GabboliniEdoardo D'AmicoCesare BernardisPaolo CremonesiPublished in: SAC (2021)
Keyphrases
- matrix factorization
- recommender systems
- collaborative filtering
- low rank
- data sparsity
- negative matrix factorization
- cold start problem
- factor analysis
- user preferences
- factorization methods
- stochastic gradient descent
- nonnegative matrix factorization
- probabilistic matrix factorization
- latent space
- implicit feedback
- variational bayesian
- data matrix
- missing data
- item recommendation
- factorization method
- recommendation quality
- user profiles
- rating prediction
- personalized recommendation
- user behavior
- latent factors
- text mining