AI Paper Atlas

Research map / Image generation / GANs

Joint inference and input optimization in equilibrium networks

Swaminathan Gurumurthy, Shaojie Bai, Zachary Manchester, J Zico Kolter

NeurIPS 2021 · Image generation › GANs · approach: Generative adversarial network theory · 2 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
Jointly solve the DEQ fixed point and input optimization with an augmented DEQ instead of separate Adam optimizationMDEQ decoder input optimization · CelebA 64×64Time: 7659 -> 862 msImproved
Apply JIIO at test time to the pretrained MDEQ-VAE decoderMDEQ-VAE · CelebA 64×64 reconstructionReconstruction FID: 45.81 -> 42.36Improved
Train a decoder-only MDEQ with JIIO latent optimizationMDEQ generative model · CelebA 64×64 reconstructionReconstruction FID: 45.81 -> 32.52Improved
Use 100 rather than 40 JIIO iterations (ablation)JIIO-MDEQ unsupervised inverse solver · CelebA denoising σ=0.2Median PSNR: 17.08 -> 17.86Improved
Use 100 rather than 40 JIIO iterations (ablation)JIIO-MDEQ unsupervised inverse solver · CelebA denoising σ=0.4Median PSNR: 16.03 -> 17.55Improved
Use 100 rather than 40 JIIO iterations (ablation)JIIO-MDEQ unsupervised inverse solver · CelebA inpaintingMedian PSNR: 15.88 -> 15.87No clear effect
Replace PGD adversarial training with JIIO adversarial trainingMDEQ classifier · MNIST, ε=1, PGD attackAccuracy: 96.53 ± 0.05 -> 95.74 ± 0.22No clear effect
Replace PGD adversarial training with JIIO adversarial trainingMDEQ classifier · CIFAR, ε=1, PGD attackAccuracy: 37.4 ± 0.26 -> 37.31 ± 0.67No clear effect
Replace PGD adversarial training with JIIO adversarial trainingMDEQ classifier · CIFAR10, ε=0.5, PGD attackAccuracy: 51.77 ± 0.75 -> 51.39 ± 0.56No clear effect
Use JIIO rather than PGD to generate adversarial examplesMDEQ classifier · MNIST adversarial example generationTime: 4360 -> 1401 msImproved
Use JIIO rather than Adam for meta-learning inner-loop optimizationDEQ meta-learning model · Omniglot 5-way 1-shotTime: more than 2.7x lesser timeImproved
Use Type-I Anderson rather than Type-II Anderson acceleration (ablation)JIIO reconstruction optimization · Pretrained MDEQ-VAE decoderOptimization lossMixed

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