Learning with the Weighted Trace-norm under Arbitrary Sampling Distributions

Abstract

We provide rigorous guarantees on learning with the weighted trace-norm under arbitrary sampling distributions. We show that the standard weighted trace-norm might fail when the sampling distribution is not a product distribution (i.e. when row and column indexes are not selected independently), present a corrected variant for which we establish strong learning guarantees, and demonstrate that it works better in practice. We provide guarantees when weighting by either the true or empirical sampling distribution, and suggest that even if the true distribution is known (or is uniform), weighting by the empirical distribution may be beneficial.

References in corpus (3)

• Restricted strong convexity and weighted matrix completion: Optimal bounds with noise
• Consistency of trace norm minimization
• Collaborative Filtering in a Non-Uniform World: Learning with the Weighted Trace Norm

Cited by in corpus (14)

• Noisy low-rank matrix completion with general sampling distribution
• A Max-Norm Constrained Minimization Approach to 1-Bit Matrix Completion
• Tensor Decompositions for temporal knowledge base completion
• Matrix Completion via Max-Norm Constrained Optimization
• Canonical Tensor Decomposition for Knowledge Base Completion
• Completing Any Low-rank Matrix, Provably
• Orthogonal Inductive Matrix Completion
• Dropout: Explicit Forms and Capacity Control
• Matrix reconstruction with the local max norm
• From bilinear regression to inductive matrix completion: a quasi-Bayesian analysis
• Multi-source Learning via Completion of Block-wise Overlapping Noisy Matrices
• A Rank-Corrected Procedure for Matrix Completion with Fixed Basis Coefficients
• Low-rank Matrix Completion in a General Non-orthogonal Basis
• Matrix Completion with Prior Subspace Information via Maximizing Correlation