Research map / Probabilistic and scientific modeling / Variational and generative sampling
Explaining the effects of non-convergent MCMC in the training of Energy-Based Models
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Sample generation using the same sampling time and random initialization as used during training (k'=k) instead of other sampling times | ConvNet EBM trained on CIFAR-10 with Langevin sampling · CIFAR-10 generation | FID | Helped |
| Sample generation using the same sampling time and initialization as training (k'=k) in Gaussian model | Gaussian EBM · Gaussian synthetic data | error in second-order correlations | Helped |
| Training with very short non-convergent chains (k=1) versus longer chains (k>=2) in Boltzmann machine (ablation) | Boltzmann machine inverse Ising · 2D Ising inverse problem | error in inferred couplings J | Hurt |
| Training with k=1 instead of k=50 steps in Boltzmann machine (ablation) | Boltzmann machine inverse Ising · 2D Ising inverse problem | quality of generated samples (2-point correlation error) | No clear effect |
| Gaussian model trained with non-convergent chains of length k where k is below threshold (c_aa - <x_a^2>_0)/2 | Gaussian EBM · Gaussian model learning dynamics | stationary coupling eigenvalues J | Hurt |
Closest papers by meaning
- Energy-Based Modelling for Discrete and Mixed Data via Heat Equations on Structured Spaces (NeurIPS 2024 · similarity 0.88)
- On scalable and efficient training of diffusion samplers (NeurIPS 2025 · similarity 0.87)
- Joint Learning of Energy-based Models and their Partition Function (ICML 2025 · similarity 0.87)
- Scalable Discrete Diffusion Samplers: Combinatorial Optimization and Statistical Physics (ICLR 2025 · similarity 0.86)
- Fast training and sampling of Restricted Boltzmann Machines (ICLR 2025 · similarity 0.86)
- Spherical Boltzmann machines: a solvable theory of learning and generation in energy-based models (NeurIPS 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.