2020/07/16 by Bryn Elesedy, Elesedy, Bryn, Varun Kanade +3
Computer Science · Economics, Econometrics and Finance · Physics and Astronomy · #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #I.5.1 #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Machine Learning and Data Classification #Model Reduction and Neural Networks #Sports Analytics and Performance
paper · pdf · doi:10.48550/arxiv.2007.08243
openalex publication_date 2020/07/16 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
We analyse the pruning procedure behind the lottery ticket hypothesis\narXiv:1803.03635v5, iterative magnitude pruning (IMP), when applied to linear\nmodels trained by gradient flow. We begin by presenting sufficient conditions\non the statistical structure of the features under which IMP prunes those\nfeatures that have smallest projection onto the data. Following this, we\nexplore IMP as a method for sparse estimation.\n