Papers
Topics
Authors
Recent
Search
2000 character limit reached

On Approximating Functions of the Singular Values in a Stream

Published 29 Apr 2016 in cs.DS | (1604.08679v2)

Abstract: For any real number p>0p > 0, we nearly completely characterize the space complexity of estimating ∥A∥pp=∑i=1nσip\|A\|_p^p = \sum_{i=1}^n \sigma_i^p for n×nn \times n matrices AA in which each row and each column has O(1)O(1) non-zero entries and whose entries are presented one at a time in a data stream model. Here the σi\sigma_i are the singular values of AA, and when p≥1p \geq 1, ∥A∥pp\|A\|_p^p is the pp-th power of the Schatten pp-norm. We show that when pp is not an even integer, to obtain a (1+ϵ)(1+\epsilon)-approximation to ∥A∥pp\|A\|_p^p with constant probability, any $1$-pass algorithm requires n1−g(ϵ)n^{1-g(\epsilon)} bits of space, where g(ϵ)→0g(\epsilon) \rightarrow 0 as ϵ→0\epsilon \rightarrow 0 and ϵ>0\epsilon > 0 is a constant independent of nn. However, when pp is an even integer, we give an upper bound of n1−2/ppoly(ϵ−1log⁡n)n^{1-2/p} \textrm{poly}(\epsilon^{-1}\log n) bits of space, which holds even in the turnstile data stream model. The latter is optimal up to poly(ϵ−1log⁡n)\textrm{poly}(\epsilon^{-1} \log n) factors. Our results considerably strengthen lower bounds in previous work for arbitrary (not necessarily sparse) matrices AA: the previous best lower bound was Ω(log⁡n)\Omega(\log n) for p∈(0,1)p\in (0,1), Ω(n1/p−1/2/log⁡n)\Omega(n^{1/p-1/2}/\log n) for p∈[1,2)p\in [1,2) and Ω(n1−2/p)\Omega(n^{1-2/p}) for p∈(2,∞)p\in (2,\infty). We note for p∈(2,∞)p \in (2, \infty), while our lower bound for even integers is the same, for other pp in this range our lower bound is n1−g(ϵ)n^{1-g(\epsilon)}, which is considerably stronger than the previous n1−2/pn^{1-2/p} for small enough constant ϵ>0\epsilon > 0. We obtain similar near-linear lower bounds for Ky-Fan norms, SVD entropy, eigenvalue shrinkers, and M-estimators, many of which could have been solvable in logarithmic space prior to our work.

Authors (2)
Citations (36)

Summary

No one has generated a summary of this paper yet.

Paper to Video (Beta)

No one has generated a video about this paper yet.

Whiteboard

No one has generated a whiteboard explanation for this paper yet.

Open Problems

We haven't generated a list of open problems mentioned in this paper yet.

Continue Learning

We haven't generated follow-up questions for this paper yet.