AI Paper Atlas

Research map / Learning theory and algorithms / Clustering

Deep Online Probability Aggregation Clustering

Yuxuan Yan, Na Lu, Ruofan Yan

ECCV 2024 · Learning theory and algorithms › Clustering · approach: Clustering approximation algorithms · 2 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 OPA self-labeling loss instead of directly minimizing J_pac (ablation)DPAC · CIFAR-100ACC (%): 50.2 -> 51.6Improved
use OPA instead of J_pac on ImageNet-Dogs (ablation)DPAC · ImageNet-DogsACC (%): 53.9 -> 65.5Improved
use weighted contrastive loss instead of standard contrastive loss (ablation)DPAC opa · CIFAR-10ACC (%): 89.7 -> 90.8Improved
use SimCLR-style framework instead of MoCo with OPA (ablation)OPA · CIFAR-10ACC (%): 86.5 -> 89.7Improved
add cluster-size regularization term to OPA (ablation)DPAC opa · CIFAR-10ACC (%): 90.7 -> 0.1Hurt
remove clustering regularization from SCAN (ablation)SCAN · CIFAR-10ACC (%): 85.7 -> 0.1Hurt
remove clustering regularization from CC (ablation)CC · CIFAR-10ACC (%): 79.2 -> 68.7Hurt
add self-labeling fine-tuning (ablation)DPAC opa · ImageNet-DogsACC (%): 65.5 -> 72.6Improved
use PAC centerless clustering vs RCC on entangled dataPAC · MNISTACC (%): 65.7 -> 59.7Hurt
use PAC instead of other clustering on pretrained deep features of subtle-class datasetsPAC · ImageNet-Dogs (deep features)ACC (%): 52.6 -> 47.3No clear effect

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