AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

A Bayesian Approach To Analysing Training Data Attribution In Deep Learning

Elisa Nguyen, Minjoon Seo, Seong Joon Oh

NeurIPS 2023 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Variational Bayesian deep learning · 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 SGD batch composition (DE-Batch) instead of initialisation (DE-Init) as the randomness source (ablation)LOO ground truth · MNIST3 CNN2-Lmean p-value of TDA: 0.331 -> 0.025Improved
use DE-Batch instead of DE-Init as randomness source (ablation)LOO ground truth · CIFAR10 CNN2-Lmean p-value of TDA: 0.692 -> 0.487Improved
use DE-Batch instead of DE-Init as randomness source (ablation)Influence functions · MNIST3 CNN2-Lmean p-value of TDA: 0.352 -> 0.000Improved
increase model complexity from CNN2-L to ViT+LoRA (ablation)LOO ground truth · MNIST3mean p-value of TDA: 0.025 -> 0.786Hurt
add a third layer (CNN2-L to CNN3-L) (ablation)LOO ground truth · MNIST3mean p-value of TDA: 0.025 -> 0.370Hurt
increase model complexity from CNN2-L to ViT+LoRA (ablation)LOO ground truth · CIFAR10mean p-value of TDA: 0.623 -> 0.777Hurt
use Grad-Cos instead of influence functionsTDA estimator · MNIST3 SWA+DE-Initmean p-value of TDA: 0.352 -> 0.003Mixed
train CNN2-L on full MNIST (60,000) rather than small subset (ablation)LOO ground truth · MNISTmean p-value of TDA: 0.331 -> 0.761Hurt
increase number of SWA/DE posterior samples (ablation)LOO ground truth · MNIST3, CIFAR10LOO p-value stabilityImproved
restrict correlation analysis to low-noise train-test pairs (LOO p<=0.05) (ablation)TDA vs LOO correlation · MNIST3, CIFAR10Spearman correlation of mean with LOOImproved
treat TDA as probabilistic (posterior mean) rather than deterministic point estimateTDA self-influence mislabel detection · MNIST3, CIFAR10fraction of mislabels foundImproved
increase training set size (ablation)TDA stability (CNN2-L) · MNIST3, CIFAR10mean p-value of TDA: MNIST3 SWA+DE-Init LOO 0.058 -> 0.714 -> 0.374 (30, 90, 180 samples)Mixed

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