AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

OpenLDN: Learning to Discover Novel Classes for Open-World Semi-Supervised Learning

Mamshad Nayeem Rizve, Navid Kardan, Salman Khan, Fahad Shahbaz Khan, Mubarak Shah

ECCV 2022 · Label-efficient and robust learning › Semi-supervised learning · approach: Safe and realistic semi-supervised learning · 49 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
add entropy regularization (ablation)OpenLDN · CIFAR-100 ablation (50% known, 50% labeled)All accuracy: 33.4 -> 53.3Improved
add pairwise similarity loss with bi-level optimization (ablation)OpenLDN · CIFAR-100 ablation (50% known, 50% labeled)Novel accuracy: 26.6 -> 40.3Improved
add closed-world SSL training (ablation)OpenLDN · CIFAR-100 ablation (50% known, 50% labeled)Novel accuracy: 40.3 -> 44.9Improved
add closed-world SSL training (ablation)OpenLDN · CIFAR-100 ablation (50% known, 50% labeled)Known accuracy: 66.2 -> 73.9Improved
add iterative pseudo-labeling (ablation)OpenLDN · CIFAR-100 ablation (50% known, 50% labeled)Novel accuracy: 44.9 -> 46.8Improved
add pairwise similarity loss (ablation)OpenLDN · CIFAR-100 ablation (50% known, 50% labeled)Known accuracy: 66.2 -> 66.2No clear effect
replace bi-level similarity with soft cosine similarity of features (ablation)OpenLDN · CIFAR-100 pairwise similarity estimation (50% known, 50% labeled)Novel accuracy: 26.6 -> 10.3Hurt
replace bi-level similarity with hard cosine threshold 0.50 (ablation)OpenLDN · CIFAR-100 pairwise similarity estimation (50% known, 50% labeled)Novel accuracy: 26.6 -> 2.1Hurt
replace bi-level similarity with hard cosine threshold 0.95 (ablation)OpenLDN · CIFAR-100 pairwise similarity estimation (50% known, 50% labeled)Novel accuracy: 26.6 -> 17.3Hurt
replace bi-level similarity with nearest neighbor similarity (ablation)OpenLDN · CIFAR-100 pairwise similarity estimation (50% known, 50% labeled)Novel accuracy: 40.3 -> 31.7Hurt
OpenLDN-UDA instead of ORCAOpenLDN-UDA · ImageNet-100Novel accuracy: 72.1 -> 68.6Hurt
OpenLDN with 10% instead of 50% labeled dataOpenLDN-MixMatch · CIFAR-100All accuracy: 38.6 -> 47.7Improved

Later papers that compared with this method

1 later accepted paper reports OpenLDN as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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