AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

$(\textrm{Implicit})^2$: Implicit Layers for Implicit Representations

Zhichun Huang, Shaojie Bai, J Zico Kolter

NeurIPS 2021 · Architectures and efficiency › RNNs and normalization · approach: Sparse dictionary representations · 1 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
replace explicit 1L-512D with implicit 1L-512D layerFourier-MFN · Image generalization, NaturalPSNR (dB): 24.57 -> 25.19Improved
replace explicit 4L-256D with implicit 4L-256D... (1L implicit vs 4L explicit comparison)Fourier-MFN · Image generalization, TextPSNR (dB): 24.84 -> 27.19Mixed
make Gabor-MFN implicitGabor-MFN · Image generalization, Text, 1L-512DPSNR (dB): 27.74 -> 28.07Improved
make SIREN implicit (iSIREN)SIREN (input inj.) · Image generalization, Natural, 1L-256DPSNR (dB): 22.88 -> 24.28Improved
make implicit models 4 layers deep (4L-256D) on image generalizationFourier-iMFN · Image generalization, NaturalPSNR (dB): 24.52 -> 24.57No clear effect
make Fourier-MFN implicit (1L-512D) for audioFourier-MFN · Audio representationPSNR (dB): 35.42 -> 43.79Improved
use deeper implicit layer (4L-256D) for audioFourier-iMFN · Audio representationPSNR (dB): 43.79 -> 47.29Improved
make Fourier-MFN implicit for video, 1L-1024DFourier-MFN · Video representationPSNR (dB): 24.97 -> 25.85Improved
make Gabor-MFN implicit for video, 4L-1024DGabor-MFN · Video representationPSNR (dB): 29.64 -> 29.20Hurt
make Fourier-MFN implicit for 3D occupancy, 1L-512DFourier-MFN · 3D occupancy, dragonIoU: 0.9650 -> 0.9698Improved
use 0th-order gradient approximation (T = 0) instead of T = 1 (ablation)Fourier-iMFN 1L-512D · Image representation 512x512training PSNRHurt
use T = 1 truncated gradient instead of T = 0 (ablation)Fourier-iMFN 1L-512D · Image representation 512x512training PSNRImproved

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