Federated Principal Component Analysis

TL;DR

This paper introduces a federated, memory-efficient, differentially private PCA algorithm that is robust to data permutation and demonstrates competitive performance and scalability in limited-memory settings.

Contribution

It presents a novel federated PCA method that is memory-efficient, permutation-invariant, and improves differential privacy guarantees over previous approaches.

Findings

Performs comparably or better than traditional PCA algorithms in limited-memory settings.

Achieves improved differential privacy with a novel covariance matrix perturbation scheme.

Exhibits good scalability and robustness in numerical simulations.

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 $\mathcal{O}(dr)$ 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 $\mathbf{X} \in \mathbb{R}^{d \times n}$ is perturbed with a non-symmetric random Gaussian matrix with variance in $\mathcal{O}\left(\left(\frac{d}{n}\right)^2 \log d \right)$, 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.