AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

SuperTickets: Drawing Task-Agnostic Lottery Tickets from Supernets via Jointly Architecture Searching and Parameter Pruning

Haoran You, Baopu Li, Zhanyi Sun, Xu Ouyang, Yingyan Lin

ECCV 2022 · Architectures and efficiency › Neural architecture search · approach: Latency-aware architecture search · 8 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 joint search-and-prune two-in-one training instead of search-then-prune (ablation)S+P (Mag.) · Cityscapes 90% sparsitymIoU (%): 42.12 -> 55.84Improved
add progressive pruning (ablation)two-in-one training · Cityscapes 90% sparsitymIoU (%): 55.84 -> 63.89Improved
add progressive pruning (ablation)two-in-one training · COCO keypoint 80% sparsityAP (%): 58.38 -> 60.14Improved
reactivate pruned weights' gradients right after pruning (IR-P) (ablation)two-in-one + PP · Cityscapes 90% sparsitymIoU (%): 63.89 -> 45.73Hurt
reactivate pruned weights' gradients during search (IR-S) (ablation)two-in-one + PP · Cityscapes 90% sparsitymIoU (%): 63.89 -> 66.61Improved
reactivate pruned weights' gradients during search (IR-S) (ablation)two-in-one + PP · COCO keypoint 80% sparsityAP (%): 60.14 -> 61.02Improved
add 50 epochs of retraining to SuperTickets (ablation)SuperTickets · Cityscapes 90% sparsitymIoU (%): 66.61 -> 67.17Improved
add 50 epochs of retraining to SuperTickets (ablation)SuperTickets · COCO keypoint 80% sparsityAP (%): 61.02 -> 61.48Improved
retrain S+P (Mag.) baselineS+P (Mag.) · Cityscapes transfer 90% sparsitymIoU (%): 42.12 -> 60.76Improved
transfer SuperTickets from ADE20K to CityscapesSuperTickets · Cityscapes 90% sparsitymIoU (%): 60.76 -> 62.91Improved
transfer ImageNet classification tickets to ADE20K segmentation vs Cityscapes ticketsSuperTickets · ADE20K 90% sparsitymIoU (%): 20.83 -> 22.42Improved
use SuperTickets instead of S+P (Mag.) at 50% pruningSuperTickets · ImageNettop-1 accuracy (%): 72.8 -> 74.2Improved

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