New method enhances MCMC sampling with learned variance reduction

Researchers have developed a new method for improving Markov chain Monte Carlo (MCMC) methods by using learned transformations to reduce variance in estimates. This approach involves training a bijection, such as a normalizing flow, to map the target distribution to a reference density in a latent space. By solving the Poisson equation in this latent space, explicit control variates can be derived, which are then transformed back to the original space. This technique also applies to importance sampling and has shown promising results in experiments on synthetic and real-world posteriors when compared to existing state-of-the-art samplers. AI

IMPACT Enhances statistical methods used in machine learning, potentially improving model training and inference.

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

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• importance sampling
• Markov chain Monte Carlo
• Normalizing Flow
• Poisson's equation
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →