AI Paper Atlas

Research map / Federated learning and privacy / Personalized and heterogeneous FL

Guiding The Last Layer in Federated Learning with Pre-Trained Models

Gwen Legate, Nicolas Bernier, Lucas Caccia, Edouard Oyallon, Eugene Belilovsky

NeurIPS 2023 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Federated learning with pretrained models · 3 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 FedNCM initialization before fine-tuning (FedNCM+FT vs FT)FT with FedAvg · CIFAR-10accuracy: 85.4 -> 87.2Improved
add FedNCM initialization before fine-tuningFT with FedAvg · Flowers-102accuracy: 64.5 -> 74.9Improved
add FedNCM initialization before fine-tuningFT with FedAvg · Stanford Carsaccuracy: 48.7 -> 54.8Improved
use FedAdam instead of FedAvg in FedNCM+FT (ablation)FedNCM+FT · CIFAR-10accuracy: 87.2 -> 89.4Improved
replace FedNCM stage 1 by 5 rounds of LP (ablation)FedNCM+FT · Flowers-102accuracy: 74.9 -> 68.6Hurt
replace FedNCM stage 1 by 10 rounds of LP (ablation)FedNCM+FT · CIFAR-10accuracy: 87.2 -> 84.5Hurt
replace linear head by three-layer MLP in LP (ablation)LP · CIFAR-10accuracy: 82.5 -> 84.1Improved
replace linear head by three-layer MLP in LP (ablation)LP · Flowers-102accuracy: 74.1 -> 72.5Hurt
replace linear head by three-layer MLP in LP (ablation)LP · EuroSAT-Subaccuracy: 92.6 -> 91.7Hurt
replace linear head by three-layer MLP in LP (ablation)LP · CUBaccuracy: 50.0 -> 50.3No clear effect
use native 224x224 input instead of 32x32 for CIFAR-10 (ablation)FedNCM+FT · CIFAR-10accuracy: 67.9 -> 87.2Improved
FedNCM instead of FT with only 90 CIFAR-10 samplesFT · CIFAR-10 small subset (ResNet18)accuracy: 45.2 -> 55.6Improved

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