AI Paper Atlas

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

Federated Data and Feature Selection by Generalized CUR Decomposition

Ying-Peng Tang, Zhuang Qi, Xiaoli Tang, Wei Zhuo, Sheng-Jun Huang, Han Yu

ICML 2026 · Federated learning and privacy › Personalized and heterogeneous FL · approach: Federated learning with pretrained models

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 per-silo local row selection on top of FedCPQR (FedGCUR vs FedCPQR) on whole-matrix reconstructionFedCPQR · vision benchmarks, IID split, k=1500Frobenius reconstruction error: CIFAR-10 WRE 30.9 -> 30.7Improved
add local row selection (FedGCUR vs FedCPQR) on non-IID reconstruction of CIFAR-100 and Tiny ImageNetFedCPQR · vision benchmarks, non-IID splitFrobenius reconstruction error: Tiny-Img WRE 44.6 -> 45.1Hurt
add local row selection (FedGCUR vs FedCPQR) on downstream accuracyFedCPQR · vision benchmarks, non-IID splitaccuracy (%): Tiny-Img 73.1 -> 65.8; C100 69.3 -> 63.0Hurt
add local row selection (FedGCUR vs FedCPQR) on downstream accuracyFedCPQR · USPS, isolet, gina prior2 (IID, rank 100)accuracy (%): USPS 42.2 -> 46.9; isolet 35.7 -> 37.9Improved
use FedCPQR pivot columns instead of variance-based feature selection (using all data)federated feature selection · isolet (IID, rank 100)accuracy (%): 24.8 -> 35.7Improved
use FedCPQR instead of centralized SciPy pivoted QRcolumn pivoting · six OpenML datasets, k=10/50/100pivot exact match: True for all; angle error on order of 1e-14No clear effect
use FedCPQR/FedGCUR rather than random or task-specific selectors on CIFAR-10 accuracyfederated selection · CIFAR-10 (IID, k=1500)accuracy (%): 90.1 vs Coreset.Var. 90.7Hurt
use FedCPQR with half the rows vs all data baselines on mfeat-pixelall methods · mfeat-pixel (1022)accuracy (%): all methods 90.0±0.0 vs oracle 92.4±0.6No clear effect
FedCPQR and FedGCUR vs FedQR runtime as parties and rank growFedQR · mfeat-pixel, gina prior2runtimeImproved

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