Get Started
Research questionHow can differentially private high-dimensional LASSO stay stable with heterogeneous covariate scales without privacy-costly preprocessing?Differential privacy makes scale normalization costly, while heterogeneous covariates can make perturbations effectively anisotropic. This can destabilize estimation and reduce accuracy in high-dimensional LASSO.
Machine Learning
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Stabilizing Private LASSO under Heterogeneous Covariates via Anisotropic Objective PerturbationThe source studies objective perturbation for differentially private high-dimensional LASSO with heterogeneous covariate scales. It uses Gram-based perturbation analysis with approximate message passing and state evolution, reporting theoretical results on convergence stability, statistical efficiency, and privacy performance relative to uniform noise injection without data-dependent preprocessing.research paper · Sep 2, 2026
Related questions
How can smooth density estimation preserve accuracy under relaxed local differential privacy?How can differentially private federated learning preserve faithful explanations without sacrificing predictive utility?How can LLM pretraining avoid sudden gradient explosions when scaling to larger models?How can we measure label-preserving adversarial vulnerability separately from ordinary misclassification in high-dimensional classifiers?
Home
Topics
Search
Library