Applying Differential Privacy to Tensor Completion

10/01/2021
by   Zheng Wei, et al.
0

Tensor completion aims at filling the missing or unobserved entries based on partially observed tensors. However, utilization of the observed tensors often raises serious privacy concerns in many practical scenarios. To address this issue, we propose a solid and unified framework that contains several approaches for applying differential privacy to the two most widely used tensor decomposition methods: i) CANDECOMP/PARAFAC (CP) and ii) Tucker decompositions. For each approach, we establish a rigorous privacy guarantee and meanwhile evaluate the privacy-accuracy trade-off. Experiments on synthetic and real-world datasets demonstrate that our proposal achieves high accuracy for tensor completion while ensuring strong privacy protections.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
10/02/2021

One-Bit Matrix Completion with Differential Privacy

Matrix completion is a prevailing collaborative filtering method for rec...
research
10/31/2018

Recovery Guarantees for Quadratic Tensors with Limited Observations

We consider the tensor completion problem of predicting the missing entr...
research
11/28/2017

Tensor Completion Algorithms in Big Data Analytics

Tensor completion is a problem of filling the missing or unobserved entr...
research
05/08/2022

GOCPT: Generalized Online Canonical Polyadic Tensor Factorization and Completion

Low-rank tensor factorization or completion is well-studied and applied ...
research
08/23/2021

Influence-guided Data Augmentation for Neural Tensor Completion

How can we predict missing values in multi-dimensional data (or tensors)...
research
01/31/2021

Beyond the Signs: Nonparametric Tensor Completion via Sign Series

We consider the problem of tensor estimation from noisy observations wit...
research
06/20/2016

Online and Differentially-Private Tensor Decomposition

In this paper, we resolve many of the key algorithmic questions regardin...

Please sign up or login with your details

Forgot password? Click here to reset