AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Deep Learning Under Privileged Information Using Heteroscedastic Dropout

John Lambert, Ozan Sener, Silvio Savarese

CVPR 2018 · Architectures and efficiency › CNN design · approach: Biologically inspired CNN explanation · 95 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 bounding-box privileged information to control heteroscedastic dropout varianceVGG-16 on ImageNet · ImageNet 75Ksingle-crop top-1 accuracy: 37.85 -> 42.30Improved
Use bounding-box privileged information to control heteroscedastic dropout varianceVGG-16 on ImageNet · ImageNet 200Ksingle-crop top-1 accuracy: 55.99 -> 55.51Hurt
Use bounding-box privileged information to control heteroscedastic dropout varianceVGG-16 on ImageNet · ImageNet 600Ksingle-crop top-1 accuracy: 66.66 -> 66.77No clear effect
Use heteroscedastic dropout with privileged imagesLSTM machine translation · Multi-30K en→deBLEU: 35.5 -> 38.4Improved
Use heteroscedastic dropout with privileged imagesLSTM machine translation · Multi-30K de→enBLEU: 40.19 -> 42.4Improved
Use heteroscedastic dropout rather than a fixed-variance Gaussian dropoutImageNet classification · ImageNet 75Ksingle-crop top-1 accuracy: 38.80 -> 42.30Improved
Replace learned covariance with entries drawn from pure Gaussian noise (ablation)Heteroscedastic dropout on image classification · ImageNetaccuracy: >3.8%Hurt
Use only a small fraction of training examples with privileged information (ablation)Heteroscedastic dropout for machine translation · Multi-30KaccuracyImproved
Use only a small fraction of training examples with privileged information (ablation)Heteroscedastic dropout for image classification · ImageNettop-1 accuracyImproved
Use adaptive validation-based learning-rate decay rather than fixed-interval decay (ablation)Image classification models · ImageNetaccuracy: 5 - 10%Improved
Remove information-bottleneck regularization (ablation)Heteroscedastic dropout · Trainingloss convergenceHurt

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