AI Paper Atlas

Research map / Continual learning / Online and replay-based

PromptCCD: Learning Gaussian Mixture Prompt Pool for Continual Category Discovery

Fernando Julio Cendra, Bingchen Zhao, Kai Han

ECCV 2024 · Continual learning › Online and replay-based · approach: Unified and unsupervised continual frameworks · 12 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 Gaussian mixture prompts (top-k=5, 100 GMM samples) to GCD (ablation)PromptCCD · CIFAR100 CCDAvg. ACC All: 58.18 -> 64.17Improved
use top-k prompt selection instead of random-k prompts (ablation)PromptCCD GMP · CIFAR100 CCDAvg. ACC New: 55.68 -> 60.34Improved
use random-k prompts instead of no prompts (ablation)PromptCCD · ImageNet-100 CCDAvg. ACC All: 69.41 -> 68.30Hurt
use random-k prompts instead of no prompts (ablation)PromptCCD · CUB CCDAvg. ACC All: 54.98 -> 51.46Hurt
increase GMM samples from 100 to 200 (ablation)PromptCCD GMP · CUB CCDAvg. ACC All: 55.45 -> 51.67Hurt
use top-k 10 instead of 5 prompts (ablation)PromptCCD GMP · CIFAR100 CCDAvg. ACC All: 64.17 -> 61.03Hurt
add GMP plug-in to Grow & Merge (ablation)Grow & Merge · CIFAR100 CCDAvg. ACC All: 57.43 -> 61.14Improved
add GMP plug-in to MetaGCD (ablation)MetaGCD · CIFAR100 CCDAvg. ACC All: 55.49 -> 58.99Improved
vary prompt pool size from 5 to 200 for L2P and DualPrompt baselines (ablation)PromptCCD-B · CUB CCDAll ACC: L2P 48.59 to 51.84; DP 53.97 to 55.54No clear effect
replace fixed-size prompt pool with Gaussian mixture poolPromptCCD-B · CIFAR100 CCDcACC All: 59.60 -> 63.97Improved
add replay to GCDGCD · CIFAR100 CCDcACC All: 58.18 -> 49.93Hurt

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