New method uses controlled missingness to reduce bias in SGD

Researchers have developed a novel method called Richardson-SGD to address bias in stochastic gradient descent when dealing with missing data. The technique involves deliberately introducing additional missingness to data, which then allows for a debiasing procedure based on Richardson extrapolation. This approach effectively reduces gradient bias from linear to quadratic dependence on the missingness ratio, improving optimization and estimation accuracy for various parametric models. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a novel technique to improve the accuracy of machine learning models trained on incomplete datasets.

RANK_REASON The cluster contains an academic paper detailing a new statistical method for machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• Ferdinand Genans
• Richardson-SGD
• Stochastic Gradient Descent

• paper
• other