Title
A characterization of constant-sample testable properties
Abstract
We characterize the set of properties of Boolean-valued functions f:X ->{0,1} on a finite domain X that are testable with a constant number of samples (x,f(x)) with x drawn uniformly at random from X. Specifically, we show that a property P is testable with a constant number of samples if and only if it is (essentially) a k-part symmetric property for some constant k, where a property is k-part symmetric if there is a partition X1, horizontal ellipsis ,Xk of X such that whether f:X ->{0,1} satisfies the property is determined solely by the densities of f on X1, horizontal ellipsis ,Xk. We use this characterization to show that symmetric properties are essentially the only graph properties and affine-invariant properties that are testable with a constant number of samples and that for every constant d >= 1, monotonicity of functions on the d-dimensional hypergrid is testable with a constant number of samples.
Year
DOI
Venue
2016
10.1002/rsa.20807
RANDOM STRUCTURES & ALGORITHMS
Keywords
DocType
Volume
Property testing,sample complexity,partial symmetry
Journal
55.0
Issue
ISSN
Citations 
1.0
1042-9832
0
PageRank 
References 
Authors
0.34
0
2
Name
Order
Citations
PageRank
Eric Blais128622.49
Yuichi Yoshida246944.88