AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Improving Generalization via Scalable Neighborhood Component Analysis

Zhirong Wu, Alexei A. Efros, Stella X. Yu

ECCV 2018 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 136 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
train with NCA and augmented memory instead of parametric softmax, kNN k=1ResNet18 · ImageNet valtop-1 accuracy: 62.91 -> 67.39Improved
train with NCA instead of softmax, kNN k=30ResNet50 · ImageNet valtop-1 accuracy: 75.09 -> 76.67Improved
compare NCA kNN with softmax classifier itselfResNet50 · ImageNet valtop-1 accuracy: 76.01 -> 76.67Improved
project baseline feature to 128-d with PCAsoftmax baseline ResNet18 · ImageNet val k=1top-1 accuracy: 62.91 -> 60.43Hurt
reduce embedding size from 128 to 64 (ablation)SNCA ResNet18 · ImageNet val k=1top-1 accuracy: 67.39 -> 65.32Hurt
reduce embedding size from 128 to 32 (ablation)SNCA ResNet18 · ImageNet val k=30top-1 accuracy: 70.59 -> 68.01Hurt
increase embedding size from 128 to 256 (ablation)SNCA ResNet18 · ImageNet val k=1top-1 accuracy: 67.39 -> 67.54No clear effect
lower temperature from 0.1 to 0.05 (ablation)SNCA ResNet18 · ImageNet val k=1top-1 accuracy: 63.87 -> 67.39Improved
lower temperature from 0.05 to 0.03 (ablation)SNCA ResNet18 · ImageNet val k=1top-1 accuracy: 67.39 -> 66.98No clear effect
use temperature 0.02 (ablation)SNCA ResNet18 · ImageNet valtop-1 accuracy: N/AHurt
use NCA instead of softmax for coarse-to-fine transferResNet · CIFAR100 from CIFAR20 trainingtop-1 accuracy: 54.17 -> 62.32Improved
use NCA instead of softmax for coarse-to-fine transferResNet · ImageNet 1000 from ImageNet127 trainingtop-1 accuracy: 48.07 -> 52.75Improved

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