Skip to content

T2A — ViT / image classifier — isotropic sensor noise

Paper evidence: main.pdf · Block findings

Lemma: D2 · Stack: pytorch Nuisance key: isotropic

Production change: Small sensor / embedding noise; class semantics unchanged.

Notebook (Run All, built-in demo): t02a-vit-isotropic.ipynb

pip install matching-pmh torch
# Open the notebook and Run All

What this task achieved (headline)

Isotropic PMH on ViT-B/16: +4.29 pp mean ImageNet-C, TDI −58% at σ=0.10.

Paper preset: t2a_vit_isotropic · from pmh.benchmark.presets import get_preset

Subtasks (paper)

ImageNet ViT-B/16 + isotropic PMH (Type 2A)

PMH ≈ ERM clean; +4.29 pp mean ImageNet-C; TDI −58% at σ=0.10.

Preset: t2a_vit_isotropic

TDI / Jacobian probes (label-free)

Layer-averaged CLS displacement under input Gaussian.

ImageNet-C transfer (15 corruptions, severity 3)

Train on Gaussian only; largest gains on noise/frost/blur.

Run with matching-pmh

from pmh import PMHTrainer, evaluate_robust_fit
# nuisance="isotropic"

Do not use PMH when

Large domain shift without D2 setup — use T4 instead.

Replace demo data with yours

Swap demo loaders for your train_loader, source_batches, target_batches, and deploy holdout. Hook the backbone before your task head.

← All 13 tasks · Quickstart