AI Paper Atlas

Research map / Label-efficient and robust learning / Noisy labels

Distributional Consistency Loss: Beyond Pointwise Data Terms in Inverse Problems

George Webber, Andrew J. Reader

ICLR 2026 · Label-efficient and robust learning › Noisy labels · approach: Contrastive learning on noisy data

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 MSE with distributional consistency loss in DIP denoisingDeep Image Prior · DIP Gaussian denoising (sigma 75/255)PSNR over iterationsImproved
replace MSE with DC loss in DIP, compared against optimally early-stopped MSEDeep Image Prior · DIP denoising across noise levelspeak PSNRImproved
replace NLL with DC loss in Adam-based PET reconstructionPET reconstruction (BrainWeb phantom) · 2D PET, Poisson noise, 10,000 iterationsimage error over iterationsImproved
compare DC-Adam with MLEM at minimum errorPET reconstruction · 2D PET, Poisson noiseminimum errorHurt
add TV regularization to DC loss versus NLL+TV (ablation)PET reconstruction, 4x lower counts · 2D PET with edge-preserving TVNRMSEImproved
use DC loss so optimal TV strength drops (ablation)PET with TV · 2D PET with edge-preserving TVoptimal regularization strength betaImproved
apply DC loss to clinical 3D PET dataReal PET brain scan (Siemens Biograph mMR) · real 3D PET, 10,000 iterations, 21 subsetslate-iteration reconstruction qualityImproved
naive PIT for discrete noise instead of randomized PIT (ablation)DC loss for Poisson noise · PET reconstructionpracticalityNo clear effect

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