Noisy Gradient Descent Converges to Flat Minima for Nonconvex Matrix Factorization.
Tianyi LiuYan LiSong WeiEnlu ZhouTuo ZhaoPublished in: AISTATS (2021)
Keyphrases
- matrix factorization
- stochastic gradient descent
- missing data
- collaborative filtering
- objective function
- low rank
- recommender systems
- cost function
- low rank matrices
- negative matrix factorization
- factor analysis
- convex optimization
- factorization methods
- variational bayesian
- data sparsity
- tensor factorization
- probabilistic matrix factorization
- incomplete data
- loss function
- factorization method
- nonnegative matrix factorization
- missing values
- data matrix
- matrix completion
- latent factor models
- personalized ranking