Max-stable sketches: estimation of Lp-norms, dominance norms and point queries for non-negative signals

TL;DR

This paper introduces max-stable sketches for efficiently estimating Lp-norms, dominance norms, and point queries on non-negative streaming data, with theoretical guarantees and practical implementation insights.

Contribution

It presents a novel approach connecting max-stable sketches to p-stable sketches, providing rigorous performance guarantees for streaming data analysis.

Findings

Efficient estimation of Lp-norms and point queries using max-stable sketches.

Theoretical error-probability bounds are established.

Algorithmic implementation details are discussed.

Abstract

Max-stable random sketches can be computed efficiently on fast streaming positive data sets by using only sequential access to the data. They can be used to answer point and Lp-norm queries for the signal. There is an intriguing connection between the so-called p-stable (or sum-stable) and the max-stable sketches. Rigorous performance guarantees through error-probability estimates are derived and the algorithmic implementation is discussed.