AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Relation-Guided Adversarial Learning for Data-Free Knowledge Transfer

Yingping Liang, Ying Fu

IJCV 2024 · Label-efficient and robust learning › Knowledge distillation · approach: Few-shot and metric distillation · 10 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
add L_ntri loss (ablation)RGAL (resnet18) · CIFAR-10top-1 accuracy: 94.07 -> 95.08Improved
add L_emb loss (ablation)RGAL (resnet18) · CIFAR-10top-1 accuracy: 93.98 -> 95.08Improved
add L_tri loss (ablation)RGAL (resnet18) · CIFAR-10top-1 accuracy: 94.43 -> 95.08Improved
add L_ntri loss (ablation)RGAL (wrn16-1) · CIFAR-100top-1 accuracy: 57.45 -> 58.73Improved
use paired batch sampling instead of random (ablation)RGAL (resnet18) · CIFAR-100top-1 accuracy: 76.91 -> 77.01Improved
use focal weighted negative sampling instead of random (ablation)RGAL (resnet18) · CIFAR-10top-1 accuracy: 94.23 -> 95.08Improved
use focal weighted instead of distance weighted negative sampling (ablation)RGAL (wrn16-1) · CIFAR-100top-1 accuracy: 58.03 -> 58.73Improved
use global embeddings instead of logits for triplet loss (ablation)RGAL (resnet18) · CIFAR-100top-1 accuracy: 76.87 -> 77.01Improved
combine logits and global embeddings (ablation)RGAL (resnet18) · CIFAR-10top-1 accuracy: 95.08 -> 95.02No clear effect
set lambda 0.5 instead of 0.7 (ablation)RGAL (resnet18) · CIFAR-10top-1 accuracy: 92.04 -> 95.08Improved
replace CMI contrastive loss with RGAL triplet losses, no pre-synthesized dataCMI (resnet18) · CIFAR-10top-1 accuracy: 93.94 -> 95.01Improved
add RGAL losses to GDFQGDFQ (W3A3) · CIFAR-100top-1 accuracy: 43.87 -> 47.53Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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