AI Paper Atlas

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

Acceleration of Federated Learning with Alleviated Forgetting in Local Training

Chencheng Xu, Zhiwei Hong, Minlie Huang, Tao Jiang

ICLR 2022 · Federated learning and privacy › Communication-efficient FL · approach: Local drift correction · 15 citations (OpenAlex)

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
Regularize local training with pseudo and perturbed data to alleviate forgettingFedReg · MNIST (two classes)R_0.9 rounds: 37 -> 25Improved
Regularize local training with pseudo and perturbed data to alleviate forgettingFedReg · MNIST (one class)R_0.9 rounds: 74 -> 32Improved
Regularize local training with pseudo and perturbed data to alleviate forgettingFedReg · EMNISTR_0.9 rounds: 170 -> 24Improved
Regularize local training with pseudo and perturbed data to alleviate forgettingFedReg · CIFAR-10 (uniform)accuracy: 0.576 -> 0.715Improved
Use pseudo-data regularization under one-class client heterogeneityFedReg · CIFAR-10 (one class)accuracy: 0.371 -> 0.616Improved
Use pseudo-data regularization with a ResNet-9FedReg · CIFAR-100 (ResNet-9)accuracy: 0.323 -> 0.502Improved
Use pseudo-data regularization while fine-tuning a transformerFedReg · CIFAR-100 (Transformer)R_0.9 rounds: 21 -> 14Improved
Use pseudo-data regularization for medical image classificationFedReg · COVID-19 CT imagesaccuracy: 0.614 -> 0.673Improved
Use modified gradients rather than DPSGD for privacy protectionFedReg with MG · EMNISTaccuracy: 0.882 -> 0.986Improved
Use modified gradients rather than DPSGD for privacy protectionFedReg with MG · CIFAR-10 (uniform)accuracy: 0.391 -> 0.703Improved
Increase pseudo-data generation step size from too-small values to a tuned intermediate valueFedReg · EMNIST and CIFAR-10 (one class)communication rounds to target accuracyMixed
Add perturbed data regularization alongside pseudo dataFedReg · Local training across evaluated datasetsaccuracyImproved

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