AI Paper Atlas

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

Regularizing Neural Networks with Meta-Learning Generative Models

Shin'ya Yamaguchi, Daiki Chijiwa, Sekitoshi Kanai, Atsutoshi Kumagai, Hisashi Kashima

NeurIPS 2023 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Normalization and unsupervised meta-learning · 1 citations (OpenAlex)

Read the paper

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
replace cross-entropy on synthetic samples with pseudo consistency regularization (PCR) (ablation)GDA · Carstop-1 accuracy (%): 84.50 -> 86.36Improved
add meta pseudo sampling (MPS) to PCR (ablation)PCR / MGR · Carstop-1 accuracy (%): 86.32 -> 87.22Improved
add MPS to plain GDA (ablation)GDA · Carstop-1 accuracy (%): 84.50 -> 85.28Improved
optimize generator G_Phi instead of the finder F_Phi (ablation)MPS · Carstop-1 accuracy (%): 87.22 -> 84.47Hurt
use MLP finder instead of residual tanh finder (ablation)MPS finder · Carstop-1 accuracy (%): 87.22 -> 86.35Hurt
add KL penalty on finder outputs (ablation)MPS · Carstop-1 accuracy (%): 86.92 -> 87.22Improved
use 1st-order finite-difference approximation instead of 2nd-order gradients (ablation)MPS · Carstop-1 accuracy (%): 87.30 -> 87.22; 6.55 -> 5.79 hoursNo clear effect
use a better generator (FastGAN to StyleGAN-XL) (ablation)MGR · Carstop-1 accuracy (%): 86.30 -> 88.37Improved
use synthetic samples instead of real samples for consistency regularization (ablation)MGR · Carstop-1 accuracy (%): 86.16 -> 87.22Improved
combine MGR with TrivialAugmentTrivialAugment base · Carstop-1 accuracy (%): 87.83 -> 89.10Improved
train finder by maximizing CE loss instead of meta-learning (ablation)MPS · Carstop-1 accuracy (%): 87.22 -> 86.52No clear effect
add LatentAugment on top of RandAugment (ablation)MGR · Carstop-1 accuracy (%): 87.22 -> 87.85Improved

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