Reproduce “Learning Parametric Distributions from Samples and Preferences” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Marc Jourdan, Gizem Yüce, Nicolas Flammarion
Researchers provide a rigorous statistical analysis demonstrating that deterministic preference feedback in parameter estimation can achieve an accelerated O(1/n) convergence rate for continuous parametric distributions, fundamentally outperforming the standard O(1/√n) parametric rate. The work includes minimax lower bounds establishing this acceleration as a statistical limit, with empirical validation across various distribution families confirming the theoretical predictions.
Key insights
- Deterministic preference feedback provides fundamentally different statistical information by imposing shrinking hard constraints on the parameter space.
- This unique information allows for accelerated statistical convergence rates beyond the traditional parametric limits.
- The O(1/n) rate achieved by deterministic preference-based estimators is minimax optimal, confirmed by matching lower bounds.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.