AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Hybrid Attention-Based Prototypical Networks for Noisy Few-Shot Relation Classification

Tianyu Gao, Xu Han, Zhiyuan Liu, Maosong Sun

AAAI 2019 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Fine-tuning strategies for few-shot

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 hybrid attention (HATT) on clean dataProto · FewRel 5-way 5-shot, 0% noiseaccuracy: 89.05 -> 90.12Improved
add hybrid attention under 10% noiseProto · FewRel 10-way 5-shotaccuracy: 79.39 -> 81.09Improved
add hybrid attention under 30% noiseProto · FewRel 5-way 5-shotaccuracy: 82.45 -> 84.71Improved
add hybrid attention under 50% noiseProto · FewRel 5-way 5-shotaccuracy: 72.91 -> 76.57Improved
add hybrid attention under 50% noiseProto · FewRel 10-way 5-shotaccuracy: 61.11 -> 65.97Improved
add hybrid attention under 50% noiseProto · FewRel 10-way 10-shotaccuracy: 71.29 -> 76.42Improved
add instance-level attention only (ablation)Proto · FewRel 5-way 5-shotaccuracy: 89.05 -> 89.63Improved
add feature-level attention only (ablation)Proto · FewRel 5-way 5-shotaccuracy: 89.05 -> 89.70Improved
combine both attentions instead of instance-level only (ablation)Proto-IATT · FewRel 10-way 5-shotaccuracy: 82.16 -> 83.05Improved
combine both attentions instead of feature-level only (ablation)Proto-FATT · FewRel 10-way 5-shotaccuracy: 82.45 -> 83.05Improved
Proto-HATT vs Proto* baseline reported by Han et al.Proto-HATT · FewRel 5-way 5-shotaccuracy: 84.79 -> 90.12Improved
hybrid attention speeds convergenceProto-HATT · FewRel trainingtraining time to reduce loss 40%Improved

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