Adapting to Noise Tails in Private Linear Regression

3464734 bytes of associated research material describe differentially private tail-robust methods for linear regression. The work implements noisy clipped gradient descent for low-dimensional settings and noisy iterative hard thresholding for high-dimensional sparse models. The proposed methods are evaluated through simulations and two real datasets.

Use Cases

• Developing privacy-preserving regression algorithms based on the use of differential privacy and Huber loss.
• Analyzing high-dimensional sparse linear models based on the implementation of noisy iterative hard thresholding.
• Studying the trade-off between bias, privacy, and robustness in statistical methods based on the tunable robustification parameter.
• Benchmarking convergence rates for regression under sub-Gaussian or heavy-tailed errors based on the described non-asymptotic analysis.

Strengths

• The dataset is associated with a detailed methodological paper that explicitly characterizes convergence rates.
• The research material is openly licensed under CC-BY-4.0.
• The dataset appears on multiple platforms, indicating its importance for the field.

Limitations

• The dataset's exact row count and column structure are unknown.
• The associated files are research outputs (PDF, ZIP, DOCX) rather than a structured data table.
• The dataset's freshness is uncertain as the last update date (2026-03-20) appears to be a future date, which may be a metadata error.

Provenance

Source

Jinyuan Chang

Collection Method

Likely generated as part of research evaluating proposed statistical methods through simulations and real datasets.

Freshness

2026-03-20