[Paper Review] Testing Linear-Invariant Non-Linear Properties

Arnab Bhattacharyya, Victor Chen+2 more|ArXiv.org|Sep 14, 2008

Complexity and Algorithms in Graphs17 references4 citations

TL;DR

This paper introduces a broad class of linear-invariant non-linear properties of Boolean functions defined by forbidden patterns under linear transformations, proving that such properties are testable when the underlying pattern corresponds to a graphic matroid. The key contribution is a generalization of Green's triangle-freeness test to an infinite family of properties, using connections between 1-complexity linear systems and graphic matroids to establish local testability with one-sided error.

ABSTRACT

We consider the task of testing properties of Boolean functions that are invariant under linear transformations of the Boolean cube. Previous work in property testing, including the linearity test and the test for Reed-Muller codes, has mostly focused on such tasks for linear properties. The one exception is a test due to Green for "triangle freeness": a function $f:\cube^{n} o\cube$ satisfies this property if $f(x),f(y),f(x+y)$ do not all equal 1, for any pair $x,y\in\cube^{n}$. Here we extend this test to a more systematic study of testing for linear-invariant non-linear properties. We consider properties that are described by a single forbidden pattern (and its linear transformations), i.e., a property is given by $k$ points $v_{1},...,v_{k}\in\cube^{k}$ and $f:\cube^{n} o\cube$ satisfies the property that if for all linear maps $L:\cube^{k} o\cube^{n}$ it is the case that $f(L(v_{1})),...,f(L(v_{k}))$ do not all equal 1. We show that this property is testable if the underlying matroid specified by $v_{1},...,v_{k}$ is a graphic matroid. This extends Green's result to an infinite class of new properties. Our techniques extend those of Green and in particular we establish a link between the notion of "1-complexity linear systems" of Green and Tao, and graphic matroids, to derive the results.

Motivation & Objective

• To investigate whether linear-invariant non-linear properties of Boolean functions can be locally tested, extending beyond previously studied linear properties.
• To characterize a broad class of such properties defined by forbidden patterns under linear transformations.
• To establish testability for properties whose forbidden patterns correspond to graphic matroids, generalizing Green's triangle-freeness result.
• To explore the structural conditions under which such properties admit one-sided error testers with constant query complexity.
• To lay the foundation for a broader understanding of the role of linear invariance in enabling local testability, even for non-linear properties.

Proposed method

• Define properties via a single forbidden pattern of k points in {0,1}^k and their linear images under L: {0,1}^k → {0,1}^n, requiring that f(L(v1)),...,f(L(vk)) do not all equal 1.
• Use the matroid structure of the forbidden pattern to classify the property, focusing on cases where the matroid is graphic (i.e., representable as the cycle space of a graph).
• Leverage the theory of 1-complexity linear systems from Green and Tao to analyze the structure of the forbidden patterns and their linear invariants.
• Establish a connection between 1-complexity systems and graphic matroids to show that the corresponding property is testable with one-sided error.
• Use extremal graph theory results (e.g., non-existence of homomorphisms between graphic matroids of complete graphs) to prove lower bounds on distance to the property.
• Apply a canonical function construction based on the matroid to demonstrate that functions far from the property can be distinguished with constant query complexity.

Experimental results

Research questions

• RQ1 Under what conditions on the forbidden pattern is a linear-invariant non-linear property testable with one-sided error?
• RQ2 Can Green's triangle-freeness test be generalized to an infinite family of non-linear, linear-invariant properties?
• RQ3 What role does the matroid structure of the forbidden pattern play in determining testability?
• RQ4 How do 1-complexity linear systems relate to graphic matroids in the context of property testing?
• RQ5 Can the framework be extended to handle arbitrary forbidden patterns, not just monochromatic ones?

Key findings

• The paper proves that any property defined by a forbidden pattern whose underlying matroid is graphic is testable with one-sided error and constant query complexity.
• The class of testable properties includes Green's triangle-freeness as a special case and generalizes it to all cycle patterns C_k for k ≥ 3.
• For a graphic matroid M corresponding to C_k, the property of being M-free is testable, and the distance to the property can be lower-bounded using extremal graph-theoretic arguments.
• The paper shows that C_{k+2}-freeness is strictly stronger than C_k-freeness, implying a proper hierarchy of such properties.
• The results establish a link between 1-complexity systems and graphic matroids, enabling the analysis of non-linear, linear-invariant properties via structural combinatorics.
• The framework provides a foundation for extending testability results to arbitrary forbidden patterns and to properties over arbitrary finite ranges and fields.