AI Paper Atlas

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

Contrastive Prototypical Network with Wasserstein Confidence Penalty

Haoqing Wang, Zhi-Hong Deng

ECCV 2022 · 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 pairwise contrast (ablation)CPN (Conv4) · miniImageNet 5-way 1-shotaccuracy (%): 46.08 -> 46.96Improved
add pairwise contrast (ablation)CPN (Conv4) · miniImageNet 5-way 50-shotaccuracy (%): 74.81 -> 75.63Improved
add pairwise contrast (ablation)CPN (ResNet12) · miniImageNet 5-way 1-shotaccuracy (%): 48.80 -> 50.01Improved
add pairwise contrast (ablation)CPN (ResNet12) · miniImageNet 5-way 50-shotaccuracy (%): 80.83 -> 82.74Improved
add Wasserstein Confidence Penalty (ablation)CPN (Conv4) · miniImageNet 5-way 1-shotaccuracy (%): 46.96 -> 47.93Improved
add Wasserstein Confidence Penalty (ablation)CPN (Conv4) · miniImageNet 5-way 50-shotaccuracy (%): 75.63 -> 78.20Improved
add Wasserstein Confidence Penalty (ablation)CPN (ResNet12) · miniImageNet 5-way 1-shotaccuracy (%): 50.01 -> 53.56Improved
add Jensen-Shannon Confidence Penalty (ablation)CPN (Conv4) · miniImageNet 5-way 1-shotaccuracy (%): 46.96 -> 46.82No clear effect
add Jensen-Shannon Confidence Penalty (ablation)CPN (ResNet12) · miniImageNet 5-way 1-shotaccuracy (%): 50.01 -> 49.87Hurt
add Consistency Regularization (ablation)CPN (Conv4) · miniImageNet 5-way 50-shotaccuracy (%): 75.63 -> 75.50Hurt
use Wasserstein penalty instead of KL Confidence Penalty (ablation)CPN (Conv4) · miniImageNet 5-way 1-shotaccuracy (%): 47.22 -> 47.93Improved
use Wasserstein penalty instead of Label Smoothing (ablation)CPN (ResNet12) · miniImageNet 5-way 1-shotaccuracy (%): 50.41 -> 53.56Improved

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