Research map / Probabilistic and scientific modeling / Variational and generative sampling
Embedded-model flows: Combining the inductive biases of model-free deep learning and explicit probabilistic modeling
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 |
|---|---|---|---|
| Add a Gaussian-mixture structured layer to MAF | MAF · 8G, CKB, MNIST | negative log probability | Helped |
| Add a Gaussian-mixture structured layer to NSF | NSF · 8G, CKB, MNIST | negative log probability | Helped |
| Add gated continuity structure | MAF · BR, OU, LZ, VDP | negative log probability | Helped |
| Add gated prior embedding to the variational flow | IAF · Structured variational inference problems | negative ELBO | Helped |
| Use a Gaussian hierarchical structured layer | MAF · IRIS and DIGITS | negative log probability | Helped |
| Use the more flexible MAF-L rather than the simple hierarchical EMF | Hierarchical generative models · IRIS and DIGITS | negative log probability | Helped |
| Place the mixture structured layer at the top rather than in the middle | EMF with MAF · 8G and CKB | negative log probability | Mixed |
| Combine NSF with the mixture layer rather than use NSF alone | NSF · 8G | negative log probability: 2.832 ± 0.004 -> 2.829 ± 0.006 | Helped |
| Combine NSF with the mixture layer rather than use NSF alone | NSF · CKB | negative log probability: 3.481 ± 0.002 -> 3.480 ± 0.001 | No clear effect |
| Combine NSF with the mixture layer rather than use NSF alone | NSF · MNIST | negative log probability: 969.393 ± 0.838 -> 958.960 ± 0.695 | Helped |
| Use gated continuity structure rather than smoothness structure | GEMF-T · BR, OU, LZ, VDP | negative log probability | Mixed |
| Use a gated prior-embedding flow rather than IAF | Variational inference · BRS-c, BRB-c, LZS-c, LZB-c, VDPS-c, VDPB-c, ES, Lin-8, Tanh-8 | negative ELBO | Helped |
Closest papers by meaning
- Automatic variational inference with cascading flows (ICML 2021 · similarity 0.89)
- SurVAE Flows: Surjections to Bridge the Gap between VAEs and Flows (NeurIPS 2020 · similarity 0.88)
- Model-Informed Flows for Bayesian Inference (NeurIPS 2025 · similarity 0.87)
- Structured Flow Autoencoders: Learning Structured Probabilistic Representations with Flow Matching (ICLR 2026 · similarity 0.87)
- Normalizing Flows With Multi-Scale Autoregressive Priors (CVPR 2020 · similarity 0.86)
- Normalizing Flows are Capable Generative Models (ICML 2025 · similarity 0.85)
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.