Research map / Probabilistic and scientific modeling / Variational and generative sampling
OT-Flow: Fast and Accurate Continuous Normalizing Flows via Optimal Transport
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 OT-Flow instead of RNODE | CNF density estimation · Power | training time (h): 25.0 -> 3.1 | Helped |
| Use OT-Flow instead of RNODE | CNF density estimation · Power | MMD: 5.64e-5 -> 4.68e-5 | Helped |
| Use OT-Flow instead of RNODE | CNF density estimation · Gas | training time (h): 36.3 -> 6.1 | Helped |
| Use OT-Flow instead of RNODE | CNF density estimation · Gas | MMD: 8.03e-5 -> 2.47e-4 | Hurt |
| Use OT-Flow instead of RNODE | CNF density estimation · Hepmass | training time (h): 46.5 -> 5.2 | Helped |
| Use OT-Flow instead of RNODE | CNF density estimation · Hepmass | MMD: 1.58e-5 -> 1.58e-5 | No clear effect |
| Use OT-Flow instead of RNODE | CNF density estimation · Miniboone | training time (h): 1.4 -> 0.8 | Helped |
| Use OT-Flow instead of RNODE | CNF density estimation · Miniboone | MMD: 2.84e-4 -> 2.84e-4 | No clear effect |
| Use OT-Flow instead of RNODE | CNF density estimation · Bsds300 | training time (h): 106.6 -> 7.1 | Helped |
| Use OT-Flow instead of RNODE | CNF density estimation · Bsds300 | MMD: 1.64e-2 -> 4.24e-4 | Helped |
| Replace exact trace computation with a single-vector Hutchinson estimator (ablation) | OT-Flow · two real-data density estimation tasks | validation loss convergence | Hurt |
| Remove the HJB regularizer while using two RK4 steps (ablation) | OT-Flow · toy Gaussian mixture | flow trajectory straightness | Hurt |
Closest papers by meaning
- Building Normalizing Flows with Stochastic Interpolants (ICLR 2023 · similarity 0.91)
- Kernelised Normalising Flows (ICLR 2024 · similarity 0.88)
- Fast and unified path gradient estimators for normalizing flows (ICLR 2024 · similarity 0.87)
- Annealing Flow Generative Models Towards Sampling High-Dimensional and Multi-Modal Distributions (ICML 2025 · similarity 0.87)
- On the Universality of Volume-Preserving and Coupling-Based Normalizing Flows (ICML 2024 · similarity 0.87)
- Switched Flow Matching: Eliminating Singularities via Switching ODEs (ICML 2024 · similarity 0.87)
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.