Subspace approximation with outliers

TL;DR

This paper introduces an efficient algorithm for subspace approximation with outliers, achieving near-optimal solutions under certain error assumptions, even with a high fraction of outliers, by extending dimension reduction and sampling techniques.

Contribution

It extends dimension reduction and sampling methods to handle outliers in subspace approximation, overcoming the SSE-hardness of robust subspace recovery under specific error conditions.

Findings

Provides a polynomial-time algorithm with linear dependence on n and d.

Achieves a (1+ε)-approximation for the optimal subspace.

Works even with large outlier fractions under certain error assumptions.

Abstract

The subspace approximation problem with outliers, for given $n$ points in $d$ dimensions $x_{1},\ldots, x_{n} \in R^{d}$, an integer $1 \leq k \leq d$, and an outlier parameter $0 \leq \alpha \leq 1$, is to find a $k$-dimensional linear subspace of $R^{d}$ that minimizes the sum of squared distances to its nearest $(1-\alpha)n$ points. More generally, the $\ell_{p}$ subspace approximation problem with outliers minimizes the sum of $p$-th powers of distances instead of the sum of squared distances. Even the case of robust PCA is non-trivial, and previous work requires additional assumptions on the input. Any multiplicative approximation algorithm for the subspace approximation problem with outliers must solve the robust subspace recovery problem, a special case in which the $(1-\alpha)n$ inliers in the optimal solution are promised to lie exactly on a $k$-dimensional linear subspace. However, robust subspace recovery is Small Set Expansion (SSE)-hard. We show how to extend dimension reduction techniques and bi-criteria approximations based on sampling to the problem of subspace approximation with outliers. To get around the SSE-hardness of robust subspace recovery, we assume that the squared distance error of the optimal $k$-dimensional subspace summed over the optimal $(1-\alpha)n$ inliers is at least $\delta$ times its squared-error summed over all $n$ points, for some $0 < \delta \leq 1 - \alpha$. With this assumption, we give an efficient algorithm to find a subset of $poly(k/\epsilon) \log(1/\delta) \log\log(1/\delta)$ points whose span contains a $k$-dimensional subspace that gives a multiplicative $(1+\epsilon)$-approximation to the optimal solution. The running time of our algorithm is linear in $n$ and $d$. Interestingly, our results hold even when the fraction of outliers $\alpha$ is large, as long as the obvious condition $0 < \delta \leq 1 - \alpha$ is satisfied.