A Non-Asymptotic Moreau Envelope Theory for High-Dimensional Generalized Linear Models

Neural Information Processing Systems 

Modern machine learning models often contain more parameters than the number of training samples.