NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:4541
Title:The spiked matrix model with generative priors

The paper provides an analysis of phase transition of spiked matrix models with generative-model priors, as opposed to the sparse prior, of previous work, and it looks at spectral and approximate message algorithm for this model. Well written. Several difference with sparse priors are noted, and it's an interesting contribution to statistical-computational gaps and student-teacher modeling.