Research map / Probabilistic and scientific modeling / Variational and generative sampling
Invertible Monotone Operators for Normalizing Flows
Byeongkeun Ahn, Chiyoon Kim, Youngjoon Hong, Hyunwoo J. Kim
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| Add monotone formulation (replace residual connections) (ablation) | i-DenseNet-style flow with CPila · CIFAR-10 | bpd: 3.243 -> 3.215 | Improved |
| Add CPila activation (replace CLipSwish) (ablation) | Monotone Flow · CIFAR-10 | bpd: 3.229 -> 3.215 | Improved |
| Add monotone formulation and CPila to i-DenseNet (ablation) | i-DenseNet · CIFAR-10 | bpd: 3.252 -> 3.215 | Improved |
| Add learnable scaling to 2 residual blocks (ablation) | 2 RBs · Toy 1D | MSE: 1.6\times10^{-2} -> 5.8\times10^{-3} | Improved |
| Replace residual blocks with monotone blocks (ablation) | 2 RBs w/ LS · Toy 1D | MSE: 5.8\times10^{-3} -> 7.6\times10^{-5} | Improved |
| Use inverse residual block as second layer instead of residual (ablation) | 2 RBs w/ LS · Toy 1D | MSE: 5.8\times10^{-3} -> 4.5\times10^{-3} | No clear effect |
| Monotone Flow vs Residual Flow | Monotone Flow · MNIST | bpd: 0.97 -> 0.928 | Improved |
| Monotone Flow vs i-DenseNet | Monotone Flow · CIFAR-10 | bpd: 3.25 -> 3.215 | Improved |
| Monotone Flow vs i-DenseNet | Monotone Flow · ImageNet32 | bpd: 3.98 -> 3.961 | Improved |
| Monotone Flow vs Residual Flow | Monotone Flow · ImageNet64 | bpd: 3.76 -> 3.734 | Improved |
| Add variational dequantization | Monotone Flow · CIFAR-10 | bpd: 3.215 -> 3.062 | Improved |
| Add variational dequantization | Monotone Flow · ImageNet32 | bpd: 3.961 -> 3.901 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Glow — Glow: Generative Flow with Invertible 1x1 Convolutions (NeurIPS 2018) · better on 4 of 4 shared results
- FFJORD — FFJORD: Free-Form Continuous Dynamics for Scalable Reversible Generative Models (ICLR 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Invertible Convolutional Flow (NeurIPS 2019 · similarity 0.87)
- CDFlow: Building Invertible Layers with Circulant and Diagonal Matrices (NeurIPS 2025 · similarity 0.87)
- Kernelised Normalising Flows (ICLR 2024 · similarity 0.86)
- Fast and unified path gradient estimators for normalizing flows (ICLR 2024 · similarity 0.86)
- Normalizing Flows are Capable Generative Models (ICML 2025 · similarity 0.85)
- On the Universality of Volume-Preserving and Coupling-Based Normalizing Flows (ICML 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.
