Meta Learning Control Variates for Variance Reduction with Limited Data

Explore an innovative approach to variance reduction in Monte Carlo estimators through a 25-minute oral presentation from the Uncertainty in Artificial Intelligence conference. Delve into the concept of Meta Learning Control Variates (Meta-CVs), a powerful technique designed to improve performance when computing multiple related integrals with limited samples. Discover how this method leverages similarities between integration tasks to achieve significant variance reduction, even in scenarios with very small sample sizes. Gain insights into the empirical assessment and theoretical analysis that establish the conditions for successful Meta-CVs implementation. Access the presentation slides to visualize key concepts and findings from this cutting-edge research in artificial intelligence and statistical computing.