AI Paper Atlas

Research map / Federated learning and privacy / Communication-efficient FL

An Efficient Framework for Clustered Federated Learning

Avishek Ghosh, Jichan Chung, Dong Yin, Kannan Ramchandran

NeurIPS 2020 · Federated learning and privacy › Communication-efficient FL · approach: Client participation analysis

Read the paper

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 IFCA cluster-specific models instead of a single global modelIFCA · Rotated MNIST (m=4800, n=50)test accuracy (%): 86.74 -> 94.20Improved
use IFCA instead of a single global modelIFCA · Rotated MNIST (m=1200, n=200)test accuracy (%): 89.73 -> 95.25Improved
use IFCA instead of a single global modelIFCA · Rotated CIFAR (m=200, n=500)test accuracy (%): 77.87 -> 81.51Improved
use IFCA instead of purely local trainingIFCA · Rotated CIFAR (m=200, n=500)test accuracy (%): 33.97 -> 81.51Improved
use IFCA instead of purely local trainingIFCA · Rotated MNIST (m=4800, n=50)test accuracy (%): 63.32 -> 94.20Improved
more samples per machine n (fewer machines m), with mn fixedlocal model baseline · Rotated MNISTtest accuracy (%): 63.32 -> 80.05Improved
more samples per machine n (fewer machines m), with mn fixedIFCA · Rotated MNISTtest accuracy (%): 94.20 -> 95.25Improved
use IFCA with weight sharing instead of global modelIFCA · FEMNISTtest accuracy (%): 84.45 -> 87.99Improved
use IFCA with weight sharing instead of local modelIFCA · FEMNISTtest accuracy (%): 75.85 -> 87.99Improved
replace one-shot centralized clustering with iterative IFCA (k=2)IFCA · FEMNISTtest accuracy (%): 87.41 -> 87.99No clear effect
replace one-shot clustering with IFCA (k=3)IFCA · FEMNISTtest accuracy (%): 87.38 -> 87.89No clear effect
increase number of clusters k from 2 to 3 (ablation)IFCA · FEMNISTtest accuracy (%): 87.99 -> 87.89No clear effect

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