33
2

Recovering PCA from Hybrid-(1,2)(\ell_1,\ell_2) Sparse Sampling of Data Elements

Abstract

This paper addresses how well we can recover a data matrix when only given a few of its elements. We present a randomized algorithm that element-wise sparsifies the data, retaining only a few its elements. Our new algorithm independently samples the data using sampling probabilities that depend on both the squares (2\ell_2 sampling) and absolute values (1\ell_1 sampling) of the entries. We prove that the hybrid algorithm recovers a near-PCA reconstruction of the data from a sublinear sample-size: hybrid-(1,2\ell_1,\ell_2) inherits the 2\ell_2-ability to sample the important elements as well as the regularization properties of 1\ell_1 sampling, and gives strictly better performance than either 1\ell_1 or 2\ell_2 on their own. We also give a one-pass version of our algorithm and show experiments to corroborate the theory.

View on arXiv
Comments on this paper