AI Paper Atlas

Research map / Image generation / Super-resolution

Controlling Neural Networks via Energy Dissipation

Michael Moeller, Thomas Mollenhoff, Daniel Cremers

ICCV 2019 · Image generation › Super-resolution · approach: CNN and transformer SR architectures · 15 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
constrain network output to be a descent direction with line search, instead of a plain baseline networkDnCNN-style SR network · BSD100 4x super-resolutionPSNR: 26.67 -> 27.12Improved
constrain network output to be a descent direction with line search, instead of a plain baseline networkDnCNN-style SR network · Set5 4x super-resolutionPSNR: 29.94 -> 31.16Improved
constrain network output to be a descent direction with line search, instead of a plain baseline networkDnCNN-style SR network · Urban100 4x super-resolutionPSNR: 24.44 -> 24.83Improved
constrain network output to be a descent direction with line search, instead of a plain baseline networkDnCNN-style SR network · BSD100 4x super-resolutiondata-fidelity error: 2.8 · 10^-1 -> 9.8 · 10^-3Improved
constrain network output to be a descent direction with line search, instead of a plain baseline networkDnCNN-style SR network · Set14 4x super-resolutionSSIM: 0.7642 -> 0.7709Improved
replace gradient descent with learned energy dissipating descentgradient descent · BSD100 4x super-resolutionPSNR: 25.09 -> 27.12Improved
use pretrained DnCNN-B denoising initialization (ablation)baseline network vs energy dissipating network · 4x super-resolutionPSNRMixed
replace baseline network with energy dissipating networkSudoku solver · 50 easy Sudokusaccuracy: 82.6% -> 87.1%Improved
replace baseline network with energy dissipating networkSudoku solver · 50 easy Sudokussolve rate: 4.0% -> 52.0%Improved
replace gradient descent with energy dissipating network, 100 iterationsSudoku solver · 50 easy Sudokussolve rate: 2.0% -> 52.0%Improved
use lagged updating of the training distribution with the network's own paths (ablation)energy dissipating network · 2D toy examplequalitative pathImproved

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