AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Ensemble Knowledge Guided Sub-network Search and Fine-Tuning for Filter Pruning

Seunghyun Lee, Byung Cheol Song

ECCV 2022 · Architectures and efficiency › Pruning and efficient tuning · approach: Structured network pruning · 10 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 single (pre-trained) teacher knowledge to the search reward (ablation)EKG search, ResNet-56 · CIFAR10accuracy (%): 93.78 -> 93.54Hurt
add single (pre-trained) teacher knowledge to the search reward (ablation)EKG search, ResNet-56 · CIFAR100accuracy (%): 71.63 -> 71.66No clear effect
add single teacher to search reward (ablation)EKG search, MobileNet-v2 · CIFAR100accuracy (%): 74.27 -> 74.10Hurt
use ensemble of interim sub-networks instead of single teacher in the search reward (ablation)EKG search, ResNet-56 · CIFAR10accuracy (%): 93.54 -> 93.85Improved
add ensemble knowledge to the search reward (vs none) (ablation)EKG search, MobileNet-v2 · CIFAR10accuracy (%): 93.69 -> 93.89Improved
add single-teacher fine-tuning guidance after ensemble search (ablation)EKG, ResNet-56 · CIFAR100accuracy (%): 71.82 -> 72.62Improved
use ensemble memory bank instead of single teacher in fine-tuning (ablation)EKG, ResNet-56 · CIFAR100accuracy (%): 72.62 -> 72.93Improved
use ensemble memory bank instead of single teacher in fine-tuning (ablation)EKG, MobileNet-v2 · CIFAR100accuracy (%): 76.11 -> 76.29Improved
use ensemble memory bank instead of single teacher in fine-tuning (ablation)EKG, ResNet-56 · fine-tuning training timetraining time increase: 71.09% -> 29.03% training time increaseImproved
use EKG score (condition-number-aware search) vs GBN scoreEKG search, ResNet-56 · CIFAR10accuracy (%): 93.70 -> 93.85Improved
EKG vs GBN despite higher validation lossEKG search, ResNet-56 · CIFAR100accuracy (%): 0.6816 -> 0.7733 validation loss; 71.60 -> 71.82 test accuracyImproved
plug EKG into Autoslim greedy searchAutoslim, ResNet-50 · ImageNet-2012top-1 accuracy (%): +0.34% top-1, +3.5% FLOPs reductionImproved

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