AI Paper Atlas

Research map / Continual learning / Online and replay-based

Enhancing Knowledge Transfer for Task Incremental Learning with Data-free Subnetwork

Qiang Gao, Xiaojun Shan, Yuchen Zhang, Fan Zhou

NeurIPS 2023 · Continual learning › Online and replay-based · approach: Gradient projection and plasticity preservation · 1 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
replace WSN weight-wise subnetwork with DSNWSN · CIFAR-100ACC: 72.84 -> 75.17Improved
replace HAT with DSNHAT · PMNISTACC: 97.10 -> 98.24Improved
replace WSN with DSNWSN · TinyImageNetACC: 45.96 -> 46.56Improved
replace WSN with DSNWSN · RMNISTACC: 97.32 -> 97.73Improved
add data-free replay and fine-tuning of similar task subnetwork (backward transfer)forgetting-free baselines (HAT, WSN) · CIFAR-100BWT: 0.00 -> 0.02Improved
add neuron-wise masks and knowledge transfer from similar tasksWSN · CIFAR-100Trans: -11.00 -> -8.67Improved
network expansion (PGN, RCL) versus fixed network with regularization (EWC)PGN/RCL · CIFAR-100ACC: 67.15 -> 53.84Hurt
mean-IMM versus naive SGDmean-IMM · PMNISTACC: 81.37 -> 80.10Hurt
larger over-parameterized hypernetwork (more neurons per layer) (ablation)DSN · RMNIST/CIFAR-100/TinyImageNetaccuracyImproved
smaller initial learning rate for hypernetwork (ablation)DSN · RMNIST/CIFAR-100/TinyImageNetaccuracyImproved
increase eta (neuron-count penalty) to reserve room for future tasks (ablation)DSN · PMNIST/TinyImageNetaccuracyHurt
set eta = 0 (no penalty on activated neurons) (ablation)DSN · PMNIST/TinyImageNetaccuracy and capacityHurt

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