Federated Principal Component Analysis

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.

[1]  Yann LeCun,et al.  The mnist database of handwritten digits , 2005 .

[2]  Anand D. Sarwate,et al.  A near-optimal algorithm for differentially-private principal components , 2012, J. Mach. Learn. Res..

[3]  Radim Rehurek Subspace Tracking for Latent Semantic Analysis , 2011, ECIR.

[4]  Thierry Bouwmans,et al.  Robust PCA via Principal Component Pursuit: A review for a comparative evaluation in video surveillance , 2014, Comput. Vis. Image Underst..

[5]  Karl Pearson F.R.S. LIII. On lines and planes of closest fit to systems of points in space , 1901 .

[6]  Jeff M. Phillips,et al.  Improved Practical Matrix Sketching with Guarantees , 2014, IEEE Transactions on Knowledge and Data Engineering.

[7]  Geoffrey E. Hinton,et al.  Visualizing Data using t-SNE , 2008 .

[8]  Nathan Srebro,et al.  Stochastic optimization for PCA and PLS , 2012, 2012 50th Annual Allerton Conference on Communication, Control, and Computing (Allerton).

[9]  Moritz Hardt,et al.  The Noisy Power Method: A Meta Algorithm with Applications , 2013, NIPS.

[10]  Zhihua Zhang,et al.  Robust Frequent Directions with Application in Online Learning , 2017, J. Mach. Learn. Res..

[11]  Charles R. Johnson,et al.  Topics in Matrix Analysis , 1991 .

[12]  Ioannis Mitliagkas,et al.  Memory Limited, Streaming PCA , 2013, NIPS.

[13]  Stephen J. Wright,et al.  On GROUSE and incremental SVD , 2013, 2013 5th IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP).

[14]  Anna Förster,et al.  Optimized Clustering Algorithms for Large Wireless Sensor Networks: A Review , 2019, Sensors.

[15]  Ameet Talwalkar,et al.  Federated Multi-Task Learning , 2017, NIPS.

[16]  Paulo Cortez,et al.  Modeling wine preferences by data mining from physicochemical properties , 2009, Decis. Support Syst..

[17]  Leland McInnes,et al.  UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction , 2018, ArXiv.

[18]  Bin Yu Assouad, Fano, and Le Cam , 1997 .

[19]  Jeff M. Phillips,et al.  Improved Practical Matrix Sketching with Guarantees , 2016, IEEE Trans. Knowl. Data Eng..

[20]  Edo Liberty,et al.  Simple and deterministic matrix sketching , 2012, KDD.

[21]  Aditya Ramamoorthy,et al.  Federated Over-the-Air Subspace Learning from Incomplete Data , 2020, ArXiv.

[22]  Blaise Agüera y Arcas,et al.  Communication-Efficient Learning of Deep Networks from Decentralized Data , 2016, AISTATS.

[23]  M. A. Iwen,et al.  A Distributed and Incremental SVD Algorithm for Agglomerative Data Analysis on Large Networks , 2016, SIAM J. Matrix Anal. Appl..

[24]  Han Liu,et al.  Minimax-Optimal Privacy-Preserving Sparse PCA in Distributed Systems , 2018, AISTATS.

[25]  Armin Eftekhari,et al.  MOSES: A Streaming Algorithm for Linear Dimensionality Reduction , 2018, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[26]  David P. Woodruff,et al.  Improved Distributed Principal Component Analysis , 2014, NIPS.

[27]  Christos Boutsidis,et al.  Online Principal Components Analysis , 2015, SODA.

[28]  David P. Woodruff,et al.  Frequent Directions: Simple and Deterministic Matrix Sketching , 2015, SIAM J. Comput..

[29]  Santosh S. Vempala,et al.  Principal Component Analysis and Higher Correlations for Distributed Data , 2013, COLT.

[30]  David P. Woodruff Sketching as a Tool for Numerical Linear Algebra , 2014, Found. Trends Theor. Comput. Sci..

[31]  Jeff M. Phillips,et al.  Relative Errors for Deterministic Low-Rank Matrix Approximations , 2013, SODA.

[32]  Cynthia Dwork,et al.  Calibrating Noise to Sensitivity in Private Data Analysis , 2006, TCC.

[33]  Cynthia Dwork,et al.  Practical privacy: the SuLQ framework , 2005, PODS.

[34]  C. Eckart,et al.  The approximation of one matrix by another of lower rank , 1936 .

[35]  Ioannis Mitliagkas Streaming PCA with Many Missing Entries , 2014 .

[36]  Aaron Roth,et al.  Beating randomized response on incoherent matrices , 2011, STOC '12.

[37]  Heng Tao Shen,et al.  Principal Component Analysis , 2009, Encyclopedia of Biometrics.

[38]  Andrew S. Tanenbaum,et al.  Distributed systems: Principles and Paradigms , 2001 .

[39]  L. Mirsky SYMMETRIC GAUGE FUNCTIONS AND UNITARILY INVARIANT NORMS , 1960 .

[40]  Raman Arora,et al.  Stochastic Optimization for Multiview Representation Learning using Partial Least Squares , 2016, ICML.

[41]  Aaron Roth,et al.  Beyond worst-case analysis in private singular vector computation , 2012, STOC '13.

[42]  N. Samatova,et al.  Principal Component Analysis for Dimension Reduction in Massive Distributed Data Sets ∗ , 2002 .

[43]  Raman Arora,et al.  Streaming Principal Component Analysis in Noisy Settings , 2018, ICML.

[44]  Tassilo Klein,et al.  Differentially Private Federated Learning: A Client Level Perspective , 2017, ArXiv.

[45]  Larry A. Wasserman,et al.  Differential privacy with compression , 2009, 2009 IEEE International Symposium on Information Theory.

[46]  Peter Richtárik,et al.  Federated Learning: Strategies for Improving Communication Efficiency , 2016, ArXiv.

[47]  Yao Lu,et al.  Oblivious Neural Network Predictions via MiniONN Transformations , 2017, IACR Cryptol. ePrint Arch..

[48]  Aaron Roth,et al.  The Algorithmic Foundations of Differential Privacy , 2014, Found. Trends Theor. Comput. Sci..

[49]  Bin Yang,et al.  Projection approximation subspace tracking , 1995, IEEE Trans. Signal Process..

[50]  Michael Naehrig,et al.  CryptoNets: applying neural networks to encrypted data with high throughput and accuracy , 2016, ICML 2016.

[51]  Martin Jaggi,et al.  COLA: Decentralized Linear Learning , 2018, NeurIPS.

[52]  Payman Mohassel,et al.  SecureML: A System for Scalable Privacy-Preserving Machine Learning , 2017, 2017 IEEE Symposium on Security and Privacy (SP).

[53]  Wei Hong,et al.  Model-Driven Data Acquisition in Sensor Networks , 2004, VLDB.

[54]  Li Zhang,et al.  Analyze gauss: optimal bounds for privacy-preserving principal component analysis , 2014, STOC.

[55]  Jimeng Sun,et al.  Streaming Pattern Discovery in Multiple Time-Series , 2005, VLDB.

[56]  Anand D. Sarwate,et al.  Near-optimal Differentially Private Principal Components , 2012, NIPS.