AI Paper Atlas

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

Bag of Instances Aggregation Boosts Self-supervised Distillation

Haohang Xu, Jiemin Fang, XIAOPENG ZHANG, Lingxi Xie, Xinggang Wang, Wenrui Dai, Hongkai Xiong, Qi Tian

ICLR 2022 · Label-efficient and robust learning › Knowledge distillation · approach: Few-shot and metric distillation · 5 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 teacher-guided bags of similar instances with bag-aggregation distillationResNet-18 and ResNet-34 students · ImageNet linear evaluationtop-1 accuracy: 52.5 -> 65.5; 57.4 -> 68.9Improved
Use teacher-guided bags of similar instancesBINGO student representations · ImageNet KNN validation, 10-NNaccuracy: 55.3 -> 61.0; 58.2 -> 64.9Improved
Use BINGO distillation instead of MoCo-v2 pretrainingResNet-18 · COCO detection, 1× scheduleAPbb: 31.3 -> 32.0Improved
Use BINGO distillation instead of MoCo-v2 pretrainingResNet-18 · COCO instance segmentation, 1× scheduleAPmk: 28.8 -> 29.6Improved
Use BINGO distillation instead of DisCoResNet-18 student with ResNet-50 teacher · ImageNet semi-supervised, 1% labelsaccuracy: 39.2 -> 42.8Improved
Use BINGO distillation instead of DisCoResNet-18 student with ResNet-50 teacher · ImageNet semi-supervised, 10% labelsaccuracy: 50.1 -> 57.5Improved
Use BINGO instead of DisCoResNet-18 student · CIFAR-10 linear evaluationaccuracy: 85.3 -> 86.8Improved
Use BINGO instead of DisCoResNet-18 student · CIFAR-100 linear evaluationaccuracy: 63.3 -> 66.5Improved
Form bags using a supervised-pretrained model rather than a self-supervised pretrained model (ablation)BINGO · ImageNet linear evaluation, ResNet-18top-1 accuracy: 64.0 -> 64.8Improved
Form bags using ground-truth labels rather than self-supervised teacher features (ablation)BINGO · ImageNet linear evaluation, ResNet-18top-1 accuracy: 64.0 -> 65.8Improved
Use a randomly initialized model to form bags instead of a pretrained model (ablation)BINGO · ImageNet linear evaluation, ResNet-18top-1 accuracy: 64.0 -> 46.6Hurt
Use teacher relation rather than online student relation to form bags (ablation)BINGO without pretrained teacher parameters · ImageNet linear evaluation, ResNet-18top-1 accuracy: 57.2 -> 62.2Improved

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