New method amortizes tempered sampling via corrected self-training
A new arXiv paper proposes training a neural sampler to amortize repeated sampling from un-normalized distributions, using corrected self-training to generate good training data.
Read the full story at arXiv stat.ML ↗
Timeline · 1 report
- 2026-11-18 05:00 · arXiv stat.ML
Amortizing Tempered Sampling with Corrected Self-Training