[Paper Review] Low-degree tests at large distances

Alex Samorodnitsky|ArXiv.org|Apr 16, 2006

Anomaly Detection Techniques and Applications3 citations

TL;DR

This paper introduces and analyzes low-degree tests for boolean functions that distinguish between quadratic polynomials and functions far from such polynomials, achieving nearly optimal trade-offs between query complexity and soundness. A key contribution is a tight inverse theorem for the third Gowers uniformity norm, enabling efficient distance estimation from the second-order Reed-Muller code beyond its list-decoding radius.

ABSTRACT

We define tests of boolean functions which distinguish between linear (or quadratic) polynomials, and functions which are very far, in an appropriate sense, from these polynomials. The tests have optimal or nearly optimal trade-offs between soundness and the number of queries. In particular, we show that functions with small Gowers uniformity norms behave ``randomly'' with respect to hypergraph linearity tests. A central step in our analysis of quadraticity tests is the proof of an inverse theorem for the third Gowers uniformity norm of boolean functions. The last result has also a coding theory application. It is possible to estimate efficiently the distance from the second-order Reed-Muller code on inputs lying far beyond its list-decoding radius.

Motivation & Objective

• To design low-degree tests with optimal or near-optimal trade-offs between query count and soundness for detecting functions far from low-degree polynomials.
• To analyze the behavior of boolean functions with small Gowers uniformity norms under hypergraph linearity tests, showing they behave 'randomly'.
• To establish an inverse theorem for the third Gowers uniformity norm of boolean functions, a central technical contribution.
• To apply the results to error-correcting codes, specifically enabling efficient distance estimation from the second-order Reed-Muller code beyond its list-decoding radius.
• To improve the soundness-query trade-off for degree-1 and degree-2 tests, particularly through relaxed rejection criteria and novel group-theoretic analysis.

Proposed method

• Uses hypergraph linearity tests to detect whether a boolean function is close to a low-degree polynomial, with perfect completeness.
• Applies the Gowers uniformity norms—particularly the third norm—as a measure of pseudorandomness to analyze test soundness.
• Employs a novel inverse theorem for the third Gowers uniformity norm to link small norms to structural properties of boolean functions.
• Leverages group-theoretic techniques, including homomorphism testing over p-groups and powers of Z_p, to bound the probability of false acceptance.
• Introduces a recursive matrix simplification process to reduce the number of query patterns and derive bounds on uniformity norms.
• Combines results from additive combinatorics and coding theory to show that the acceptance probability of a local test for the second-order Reed-Muller code is tightly bounded by the Gowers uniformity norm.

Experimental results

Research questions

• RQ1 What is the optimal trade-off between the number of queries and soundness in low-degree tests for boolean functions?
• RQ2 How do functions with small Gowers uniformity norms behave under hypergraph linearity tests?
• RQ3 Can an inverse theorem for the third Gowers uniformity norm be established for boolean functions?
• RQ4 To what extent can the distance from the second-order Reed-Muller code be estimated efficiently for inputs beyond the list-decoding radius?
• RQ5 What is the soundness of a relaxed degree-1 test and how does it compare to standard linearity tests?

Key findings

• The paper establishes a tight inverse theorem for the third Gowers uniformity norm of boolean functions, showing that small norms imply structural approximation by quadratic polynomials.
• The soundness of the degree-2 test is bounded by $ s riangleq rac{1}{2^{|E|}} + ext{negligible} $, where $ |E| $ is the number of queries, achieving near-optimal trade-off.
• Functions with small Gowers uniformity norms behave pseudorandomly with respect to hypergraph linearity tests, justifying their use in PCP constructions.
• It is possible to estimate the distance from the second-order Reed-Muller code efficiently even for inputs lying far beyond the list-decoding radius, due to the tight analysis of the local test.
• The soundness of the relaxed degree-1 test is asymptotically optimal, improving upon standard linearity tests by a significant margin.
• A new homomorphism testing result is proven: if a function $ heta: G \to H $ satisfies $ \Pr_{x,y} (\theta(x)+\theta(y)=\theta(x+y)) \geq \epsilon $, then there exists a homomorphism $ \psi $ such that $ \Pr_x (\theta(x) = \psi(x) + h) \geq c \cdot r^{-c'} \cdot \epsilon^{c''} $, with absolute constants $ c, c', c'' $.