AI Paper Atlas

Research map / Continual learning / Online and replay-based

Few-Shot and Continual Learning With Attentive Independent Mechanisms

Eugene Lee, Cheng-Han Huang, Chen-Yi Lee

ICCV 2021 · Continual learning › Online and replay-based · approach: Gradient projection and plasticity preservation · 32 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 AIM module to SIBSIB (Conv-4-64) · MiniImageNet 5-way 1-shotaccuracy: 58.0 -> 61.90Improved
add AIM module to SIBSIB (WRN-28-10) · MiniImageNet 5-way 1-shotaccuracy: 70.0 -> 71.22Improved
add AIM module to SIBSIB (WRN-28-10) · CIFAR-FS 5-way 5-shotaccuracy: 85.3 -> 87.34Improved
replace AIM with a parameter-matched linear layer in SIB (ablation)SIB + Linear (Conv-4-64) · MiniImageNet 5-way 1-shotaccuracy: 60.07 -> 61.90Improved
replace AIM with a parameter-matched linear layer in SIB (ablation)SIB + Linear (WRN-28-10) · CIFAR-FS 5-way 5-shotaccuracy: 79.91 -> 87.34Improved
use soft (attention above 0.5) instead of hard top-K mechanism selection (ablation)AIM (Conv-4-64) · MiniImageNet 5-way 1-shotaccuracy: 61.90 -> 61.74No clear effect
use soft (attention above 0.5) instead of hard top-K mechanism selection (ablation)AIM (WRN-28-10) · MiniImageNet 5-way 1-shotaccuracy: 71.03 -> 69.96Hurt
use soft instead of hard mechanism selection (ablation)AIM (WRN-28-10) · CIFAR-FS 5-way 1-shotaccuracy: 79.19 -> 80.14Improved
increase stochastic sampling count K+l from 10 to 32 (ablation)AIM (Conv-4-64) · CIFAR-FS 5-way 1-shotaccuracy: 71.09 -> 66.90Hurt
increase active mechanisms K from 1 to 2 (ablation)AIM (Conv-4-64) · MiniImageNet 5-way 1-shotaccuracy: 25.54 -> 62.00Improved
increase active mechanisms K from 2 to 32 (ablation)AIM (Conv-4-64) · MiniImageNet 5-way 5-shotaccuracy: 74.62 -> 74.79No clear effect
add AIM to OMLOML · Omniglot (100 classes)accuracy: 73.61 -> 97.75Improved

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