NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:1932
Title:Learning Sample-Specific Models with Low-Rank Personalized Regression

The paper introduces a new method for personalized regression, and tests it empirically on several problems. This is overall a nice contribution, the reviewers found that the paper brings a novel solution to a specific but common problem, and were overall happy with the detailed experimental study that supports the relevance of the proposed method on a variety of tasks. However, the novelty of the work is a bit limited given the extensive previous work on multitask learning and domain shift adaptation. The lack of theoretical analysis of the procedure (probably difficult due to the fact that the proposed penalty is not convex) also limits the scope of the contribution.