AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Rep-Net: Efficient On-Device Learning via Feature Reprogramming

Li Yang, Adnan Siraj Rakin, Deliang Fan

CVPR 2022 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training

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
Use Rep-Net instead of TinyTLProxylessNAS-Mobile · Aircraftaccuracy: 75.4 -> 77.4(↑2%)Improved
Use Rep-Net instead of TinyTLProxylessNAS-Mobile@320 · CUBaccuracy: 81.0 -> 82.3(↑1.3%)Improved
Use Rep-Net instead of TinyTLProxylessNAS-Mobile · Flowersaccuracy: 95.5 -> 96.1Improved
Use Rep-Net instead of TinyTLProxylessNAS-Mobile · Carsaccuracy: 85.0 -> 85.8Improved
Use Rep-Net instead of TinyTLProxylessNAS-Mobile · CUBaccuracy: 77.1 -> 77.8Improved
Use Rep-Net instead of TinyTLProxylessNAS-Mobile · Foodaccuracy: 79.7 -> 80.5Improved
Use Rep-Net instead of TinyTLProxylessNAS-Mobile · Petsaccuracy: 91.8 -> 91.8No clear effect
Use Rep-Net instead of TinyTLProxylessNAS-Mobile · CIFAR10accuracy: 95.9 -> 95.9No clear effect
Use Rep-Net instead of TinyTLProxylessNAS-Mobile · CIFAR100accuracy: 81.4 -> 81.9Improved
Add dual feature exchange so both backbone and Rep-Net features are updated (ablation)Rep-Net · CUBaccuracy: 65.5 -> 77.1Improved
Add backbone features to the connector exchange (ablation)Rep-Net · CUBaccuracy: 67.8 -> 77.1Improved
Add Rep-Net features to the connector exchange (ablation)Rep-Net · CUBaccuracy: 71.4 -> 77.1Improved

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