Research map / Probabilistic and scientific modeling / Variational and generative sampling
Implicit Normalizing Flows
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 |
|---|---|---|---|
| Use ImpFlow instead of ResFlow with the same ResNet-18-sized architecture | ResNet-18 · CIFAR10 test | classification error rate (%): 8.24 ($\pm$0.03) -> 7.29 ($\pm$0.03) | Helped |
| Use ImpFlow instead of ResFlow with the same ResNet-18-sized architecture | ResNet-18 · CIFAR100 test | classification error rate (%): 31.02 ($\pm$0.05) -> 29.06 ($\pm$0.03) | Helped |
| Use parameter-matched ImpFlow (L=5) instead of ResFlow (L=10) | normalizing flow · POWER test | average test log-likelihood (nats): 0.26 -> 0.30 | Helped |
| Use parameter-matched ImpFlow (L=5) instead of ResFlow (L=10) | normalizing flow · GAS test | average test log-likelihood (nats): 6.20 -> 6.94 | Helped |
| Use parameter-matched ImpFlow (L=5) instead of ResFlow (L=10) | normalizing flow · HEPMASS test | average test log-likelihood (nats): -18.91 -> -18.52 | Helped |
| Use parameter-matched ImpFlow (L=5) instead of ResFlow (L=10) | normalizing flow · MINIBOONE test | average test log-likelihood (nats): -21.81 -> -21.50 | Helped |
| Use parameter-matched ImpFlow (L=5) instead of ResFlow (L=10) | normalizing flow · BSDS300 test | average test log-likelihood (nats): 104.63 -> 113.72 | Helped |
| Use ImpFlow (L=20) rather than the strongest listed baseline, NAF | normalizing flow · GAS test | average test log-likelihood (nats): 11.96 -> 12.11 | Helped |
| Use ImpFlow (L=20) rather than the strongest listed baseline, FFJORD | normalizing flow · HEPMASS test | average test log-likelihood (nats): -14.92 -> -13.95 | Helped |
| Use ImpFlow (L=20) rather than the strongest listed baseline, NAF | normalizing flow · MINIBOONE test | average test log-likelihood (nats): -8.86 -> -13.32 | Hurt |
| Use ImpFlow (L=20) rather than the strongest listed baseline, NAF | normalizing flow · BSDS300 test | average test log-likelihood (nats): 157.73 -> 155.68 | Hurt |
| Use ImpFlow (L=20) rather than the strongest listed baseline, NAF | normalizing flow · POWER test | average test log-likelihood (nats): 0.62 -> 0.61 | Hurt |
Closest papers by meaning
- Self Normalizing Flows (ICML 2021 · similarity 0.88)
- Kernelised Normalising Flows (ICLR 2024 · similarity 0.86)
- On the Universality of Volume-Preserving and Coupling-Based Normalizing Flows (ICML 2024 · similarity 0.86)
- Flexible Tails for Normalizing Flows (ICML 2025 · similarity 0.85)
- Normalizing Flows are Capable Generative Models (ICML 2025 · similarity 0.85)
- Fast and unified path gradient estimators for normalizing flows (ICLR 2024 · 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.