AI Paper Atlas

Research map / Probabilistic and scientific modeling / Bayesian optimization

Bayesian Optimization with Clustering and Rollback for CNN Auto Pruning

Hanwei Fan, Jiandong Mu, Wei Zhang

ECCV 2022 · Probabilistic and scientific modeling › Bayesian optimization · approach: Multi-fidelity hyperparameter optimization · 7 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 layer clusteringNaive BO ResNet56 · CIFAR-10top-1 accuracy (%): 90.29 -> 91.68Improved
add rollback after layer clusteringLayer Clustering ResNet56 · CIFAR-10top-1 accuracy (%): 91.68 -> 92.53Improved
add layer clusteringNaive BO MobileNetV1 · ImageNettop-1 accuracy (%): 47.39 -> 48.49Improved
add rollback after layer clusteringLayer Clustering MobileNetV1 · ImageNettop-1 accuracy (%): 48.49 -> 49.34Improved
add layer clusteringNaive BO MobileNetV2 · ImageNettop-1 accuracy (%): 51.09 -> 52.57Improved
add direct rollback after layer clusteringLayer Clustering MobileNetV2 · ImageNettop-1 accuracy (%): 52.57 -> 52.86Improved
use gradual rollback instead of direct rollback (ablation)Layer Clustering (Structure) MobileNetV2 · ImageNettop-1 accuracy gain (points): +0.41 -> +1.60Improved
use gradual rollback instead of direct rollback (ablation)Layer Clustering (Euclidean) MobileNetV2 · ImageNettop-1 accuracy gain (points): +0.29 -> +0.44Improved
replace RL agent with Bayesian optimizationRL pruning agent MobileNetV2 · ImageNettop-1 accuracy (%): 43.15 -> 51.09Improved
use structure similarity instead of Euclidean distribution similarity (ablation)Layer Clustering ResNet56 · CIFAR-10top-1 accuracy (%): 88.95 -> 91.68Improved
use Euclidean distribution similarity instead of structure similarity (ablation)Layer Clustering MobileNetV2 · ImageNettop-1 accuracy (%): 50.31 -> 52.57Improved
use Euclidean instead of JS divergence similarity (ablation)Layer Clustering MobileNetV1 · ImageNettop-1 accuracy (%): 47.99 -> 48.50Improved

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