Research map / Probabilistic and scientific modeling / Variational and generative sampling
Unconstrained Monotonic Neural Networks
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 UMNN-MAF instead of B-NAF transformation | autoregressive flow · POWER | test NLL (nats): −0.61 -> −0.63 | Helped |
| Use UMNN-MAF instead of B-NAF | autoregressive flow · GAS | test NLL (nats): −12.06 -> −10.89 | Hurt |
| Use UMNN-MAF instead of B-NAF | autoregressive flow · HEPMASS | test NLL (nats): 14.71 -> 13.99 | Helped |
| Use UMNN-MAF instead of B-NAF | autoregressive flow · MINIBOONE | test NLL (nats): 8.95 -> 9.67 | Hurt |
| Use UMNN-MAF instead of B-NAF | autoregressive flow · BSDS300 | test NLL (nats): −157.36 -> −157.98 | Helped |
| Use UMNN-MAF instead of MAF (single flow, no multi-scale) | MNIST density model · MNIST density estimation | test NLL (bits/dim): 1.89 -> 1.13 | Helped |
| Add UMNN-MAF flow to VAE posterior vs plain VAE | VAE · MNIST VAE | negative ELBO (nats): 86.65 -> 84.11 | Helped |
| Use UMNN-MAF flow instead of B-NAF in VAE | VAE · MNIST VAE | negative ELBO (nats): 83.59 -> 84.11 | Hurt |
| Use UMNN-MAF flow in VAE vs plain VAE | VAE · Caltech 101 VAE | negative ELBO (nats): 110.80 -> 110.45 | No clear effect |
| Share the integrand network weights across dimensions (ablation) | UMNN-MAF · density estimation | performance | No clear effect |
Closest papers by meaning
- Constrained Monotonic Neural Networks (ICML 2023 · similarity 0.89)
- Advancing Constrained Monotonic Neural Networks: Achieving Universal Approximation Beyond Bounded Activations (ICML 2025 · similarity 0.89)
- Smooth Min-Max Monotonic Networks (ICML 2024 · similarity 0.88)
- Counterexample-Guided Learning of Monotonic Neural Networks (NeurIPS 2020 · similarity 0.88)
- Learning Monotonic Probabilities with a Generative Cost Model (ICML 2025 · similarity 0.83)
- Large Stepsize Gradient Descent for Non-Homogeneous Two-Layer Networks: Margin Improvement and Fast Optimization (NeurIPS 2024 · similarity 0.83)
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.