AI Paper Atlas

Research map / Continual learning / Online and replay-based

Continual Learning From a Stream of APIs

Enneng Yang, Zhenyi Wang, Li Shen, Nan Yin, Tongliang Liu, Guibing Guo, X. Wang, Dacheng Tao

TPAMI 2024 · Continual learning › Online and replay-based · approach: Unified and unsupervised continual frameworks · 6 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 memory replay loss (ablation)DFCL-APIs · NLP 5-dataset sequenceaccuracy (%): 58.49 -> 66.20Improved
add network similarity loss (ablation)DFCL-APIs · NLP 5-dataset sequenceaccuracy (%): 64.45 -> 66.20Improved
add both memory replay and network similarity losses (ablation)DFCL-APIs · NLP 5-dataset sequenceaccuracy (%): 54.68 -> 66.20Improved
provide 2% raw data (DECL-APIs instead of DFCL-APIs)DFCL-APIs · NLP 5-dataset sequenceaccuracy (%): 66.20 -> 72.62Improved
provide 2% raw data (DECL-APIs instead of DFCL-APIs)DFCL-APIs · Split-SVHNaccuracy (%): 94.29 -> 96.01Improved
provide 10% raw data (DECL-APIs instead of DFCL-APIs)DFCL-APIs · Split-CIFAR100accuracy (%): 30.52 -> 51.90Improved
provide 10% raw data (DECL-APIs instead of DFCL-APIs)DFCL-APIs · Split-MiniImageNetaccuracy (%): 22.95 -> 51.49Improved
use DECL-APIs with 2% data instead of Classic CL with 2% dataClassic CL · Split-SVHNaccuracy (%): 55.14 -> 96.01Improved
use DECL-APIs with 2% data instead of Classic CL with 2% dataClassic CL · NLP 5-dataset sequenceaccuracy (%): 27.96 -> 72.62Improved
return only hard labels from the API instead of soft probabilitiesDFCL-APIs · Split-MNISTaccuracy (%): 98.58 -> 95.29Hurt
return only hard labels from the API instead of soft probabilitiesDECL-APIs (2%) · Split-MNISTaccuracy (%): 98.60 -> 97.65Hurt
black-box API access instead of white-box model accessEx-Model · Split-CIFAR100accuracy (%): 36.98 -> 30.52Hurt

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