AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

On the Out-of-distribution Generalization of Probabilistic Image Modelling

Mingtian Zhang, Andi Zhang, Steven McDonagh

NeurIPS 2021 · Label-efficient and robust learning › OOD and anomaly detection · approach: Spurious-aware OOD detection · 3 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 local model instead of full PixelCNN on in-distribution dataLocal autoregressive · Fashion MNIST IDBPD: 2.78 -> 2.89Hurt
Use local model instead of full PixelCNN on in-distribution dataLocal autoregressive · CIFAR10 IDBPD: 3.12 -> 3.25Hurt
Use local model instead of full PixelCNN on OOD dataLocal autoregressive · Trained on Fashion MNIST, OOD MNIST/KMNISTBPD: 1.50 -> 1.49No clear effect
Use local model instead of full PixelCNN on OOD dataLocal autoregressive · Trained on CIFAR10, SVHN/CelebABPD: 2.13 -> 2.13No clear effect
Increase local horizon size h from 1 to 5 (ablation)Local model · Fashion MNIST IDBPD: 3.17 -> 2.88Improved
Increase local horizon size beyond h=2 (ablation)Local model · Fashion MNIST trained, OOD MNISTBPD: 1.48 -> 1.51Hurt
Increase local horizon size from h=2 to h=5 (CIFAR10) (ablation)Local model · CIFAR10 IDBPD: 3.38 -> 3.25Improved
Increase horizon size from h=2 to h=5 on OOD (CIFAR10 trained) (ablation)Local model · SVHNBPD: 2.21 -> 2.15No clear effect
Use non-local likelihood ratio score instead of full PixelCNN likelihood for OOD detectionOur method · CIFAR10 vs CelebAAUROC: 0.602 -> 0.949Improved
Use non-local likelihood ratio score instead of full PixelCNN likelihoodOur method · FashionMNIST vs MNISTAUROC: 0.074 -> 1.000Improved
Increase NeLLoC residual blocks r from 0 to 3 (trained on CIFAR) (ablation)NeLLoC · CIFAR IDBPD: 3.38 -> 3.25Improved
Split full ImageNet images into 400 patches for parallel decodingNeLLoC · Full ImageNetBPD: 3.25 -> 3.34Hurt

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