Research map / Probabilistic and scientific modeling / Variational and generative sampling
Probabilistic Unrolling: Scalable, Inverse-Free Maximum Likelihood Estimation for Latent Gaussian Models
Alexander Lin, Bahareh Tolooshams, Yves Atchade, Demba E. Ba
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Replace matrix-inversion gradient EM with probabilistic unrolling | Noisy AR parameter recovery · D=1,000 | training time: 41±0 s -> 8±0 s | Improved |
| Replace matrix-inversion gradient EM with probabilistic unrolling | Noisy AR parameter recovery · D=30,000 | training time: 4139±49 s -> 87±1 s | Improved |
| Replace matrix-inversion gradient EM with probabilistic unrolling | Noisy AR parameter recovery · D=30,000 | r(φ, φ*): 0.5±0.2 % -> 0.4±0.2 % | Improved |
| Replace matrix-inversion gradient EM with probabilistic unrolling | Bayesian compressed sensing · NIST digits average | training time: 1481±19 s -> 21±0 s | Improved |
| Replace matrix-inversion gradient EM with probabilistic unrolling | Bayesian compressed sensing · NIST digits average | NRMSE: 4.8±1.0 % -> 4.7±1.4 % | No clear effect |
| Replace matrix-inversion gradient EM with probabilistic unrolling | Factor-analysis collaborative filtering · MovieLens ML-1M | time per cycle: 54 min, 42 s -> 5 min, 50 s | Improved |
| Replace matrix-inversion gradient EM with probabilistic unrolling | Factor-analysis collaborative filtering · MovieLens ML-1M | RMSE: 0.8433 -> 0.8436 | No clear effect |
| Replace matrix-inversion gradient EM with probabilistic unrolling | Factor-analysis collaborative filtering · MovieLens ML-10M | time per cycle: 78 min, 36 s -> 12 min, 8 s | Improved |
| Replace matrix-inversion gradient EM with probabilistic unrolling | Factor-analysis collaborative filtering · MovieLens ML-10M | RMSE: 0.7809 -> 0.7796 | Improved |
| Use probabilistic unrolling to fit the factor-analysis model | Factor-analysis collaborative filtering · MovieLens ML-25M | feasibility under memory limit | Improved |
Closest papers by meaning
- Unrolled denoising networks provably learn to perform optimal Bayesian inference (NeurIPS 2024 · similarity 0.87)
- Learning Deep Latent Gaussian Models with Markov Chain Monte Carlo (ICML 2017 · similarity 0.87)
- Generic Inference in Latent Gaussian Process Models (JMLR 2019 · similarity 0.87)
- Learning Latent Variable Models via Jarzynski-adjusted Langevin Algorithm (NeurIPS 2025 · similarity 0.85)
- UNEM: UNrolled Generalized EM for Transductive Few-Shot Learning (CVPR 2025 · similarity 0.84)
- Information Theoretic Learning for Diffusion Models with Warm Start (NeurIPS 2025 · similarity 0.84)
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.
