Federated Principal Component Analysis

arXiv1907.08059MaRDI QIDQ6322324

Andreas Grammenos, Cecilia Mascolo, Jon Crowcroft, Rodrigo Mendoza-Smith

Publication date: 18 July 2019

Abstract: We present a federated, asynchronous, and $(varepsilon,delta)$-differentially private algorithm for PCA in the memory-limited setting. Our algorithm incrementally computes local model updates using a streaming procedure and adaptively estimates its $r$ leading principal components when only $mathcalO(dr)$ memory is available with $d$ being the dimensionality of the data. We guarantee differential privacy via an input-perturbation scheme in which the covariance matrix of a dataset $mathbfXinmathbb{R}^{dimesn}$ is perturbed with a non-symmetric random Gaussian matrix with variance in $mathcalOleft(left(fracdnight{)}^{2}logdight)$, thus improving upon the state-of-the-art. Furthermore, contrary to previous federated or distributed algorithms for PCA, our algorithm is also invariant to permutations in the incoming data, which provides robustness against straggler or failed nodes. Numerical simulations show that, while using limited-memory, our algorithm exhibits performance that closely matches or outperforms traditional non-federated algorithms, and in the absence of communication latency, it exhibits attractive horizontal scalability.

Has companion code repository: https://github.com/andylamp/federated_pca

This page was built for publication: Federated Principal Component Analysis

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6322324)