AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Probabilistic Deep Ordinal Regression Based on Gaussian Processes

Yanzhu Liu, Fan Wang, Adams Wai Kin Kong

ICCV 2019 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: PAC-Bayes and decoupled GP learning · 26 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
train CNN and GP jointly end-to-end instead of separately (logistic likelihood) (ablation)GP-DNNOR · Urban image aestheticsaccuracy (%): 67.37 -> 68.29Improved
train CNN and GP jointly end-to-end instead of separately (logistic likelihood) (ablation)GP-DNNOR · Urban image aestheticsMAE: 0.34 -> 0.32Improved
use logistic instead of Gaussian ordinal likelihood (joint training) (ablation)GP-DNNOR · Urban image aestheticsaccuracy (%): 67.32 -> 68.29Improved
use logistic instead of Gaussian ordinal likelihood (joint training) (ablation)GP-DNNOR · Urban image aestheticsMAE: 0.34 -> 0.32Improved
use logistic instead of Gaussian likelihood when training separately (ablation)CNN + GP · Urban image aestheticsMAE: 0.35 -> 0.34No clear effect
replace multi-class GP layer (CNNm-GP) with ordinal GP layer with learned boundariesGP-DNNOR · Urban image aestheticsMAE: 0.41 -> 0.32Improved
replace multi-class GP layer with ordinal GP layerGP-DNNOR · Adience Faceaccuracy (%): 46.1 -> 57.4Improved
replace multi-class GP layer with ordinal GP layerGP-DNNOR · Historical Color Imageaccuracy (%): 49.98 -> 46.60Hurt
replace multi-class GP layer with ordinal GP layerGP-DNNOR · Historical Color ImageMAE: 0.84 -> 0.76Improved
replace softmax classifier CNNm with ordinal GP layerGP-DNNOR · Adience FaceMAE: 0.61 -> 0.54Improved
compare GP-DNNOR with CNNPOR on accuracyGP-DNNOR · Historical Color Imageaccuracy (%): 50.12 -> 46.60Hurt

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