AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

FFJORD: Free-Form Continuous Dynamics for Scalable Reversible Generative Models

Will Grathwohl, Ricky T. Q. Chen, Jesse Bettencourt, Ilya Sutskever, David Duvenaud

ICLR 2019 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Expressive invertible architectures · 127 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use FFJORD instead of the strongest reported autoregressive baselineDensity estimation · POWERnegative log-likelihood: -0.62 -> -0.46Hurt
Use FFJORD instead of the strongest reported autoregressive baselineDensity estimation · GASnegative log-likelihood: -11.96 -> -8.59Hurt
Use FFJORD instead of the strongest reported autoregressive baselineDensity estimation · HEPMASSnegative log-likelihood: 15.09 -> 14.92Improved
Use FFJORD instead of the strongest reported autoregressive baselineDensity estimation · MINIBOONEnegative log-likelihood: 8.86 -> 10.43Hurt
Use FFJORD instead of the strongest reported autoregressive baselineDensity estimation · BSDS300negative log-likelihood: -157.73 -> -157.40Hurt
Use FFJORD multiscale instead of GlowImage density estimation · MNISTnegative log-likelihood (bits/dim): 1.05* -> 0.99*Improved
Use FFJORD multiscale instead of GlowImage density estimation · CIFAR10negative log-likelihood (bits/dim): 3.35* -> 3.40*Hurt
Use FFJORD instead of the strongest competing normalizing flowVAE approximate posterior · MNISTnegative ELBO: 83.32 ± .06 -> 82.82 ± .01Improved
Use FFJORD instead of the strongest competing normalizing flowVAE approximate posterior · Omniglotnegative ELBO: 99.00 ± .04 -> 98.33 ± .09Improved
Use FFJORD instead of the strongest competing normalizing flowVAE approximate posterior · Frey Facesnegative ELBO (bits/dim): 4.40 ± .06 -> 4.39 ± .01Improved
Use FFJORD instead of the strongest competing normalizing flowVAE approximate posterior · Caltech Silhouettesnegative ELBO: 104.62 ± .29 -> 104.03 ± .43Improved
Add the bottleneck trace-estimation trick (ablation)FFJORD training with Gaussian Hutchinson noise · MNISTtraining convergence speedImproved

Later papers that compared with this method

20 later accepted papers report FFJORD as a baseline in their results tables; 14 of them report a better number on most of the results they share with it.

12 more not listed.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most 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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy