English

Testing $k$-Monotonicity

Data Structures and Algorithms 2016-09-15 v2 Discrete Mathematics Machine Learning

Abstract

A Boolean kk-monotone function defined over a finite poset domain D{\cal D} alternates between the values 00 and 11 at most kk times on any ascending chain in D{\cal D}. Therefore, kk-monotone functions are natural generalizations of the classical monotone functions, which are the 11-monotone functions. Motivated by the recent interest in kk-monotone functions in the context of circuit complexity and learning theory, and by the central role that monotonicity testing plays in the context of property testing, we initiate a systematic study of kk-monotone functions, in the property testing model. In this model, the goal is to distinguish functions that are kk-monotone (or are close to being kk-monotone) from functions that are far from being kk-monotone. Our results include the following: - We demonstrate a separation between testing kk-monotonicity and testing monotonicity, on the hypercube domain {0,1}d\{0,1\}^d, for k3k\geq 3; - We demonstrate a separation between testing and learning on {0,1}d\{0,1\}^d, for k=ω(logd)k=\omega(\log d): testing kk-monotonicity can be performed with 2O(dlogdlog1/ε)2^{O(\sqrt d \cdot \log d\cdot \log{1/\varepsilon})} queries, while learning kk-monotone functions requires 2Ω(kd1/ε)2^{\Omega(k\cdot \sqrt d\cdot{1/\varepsilon})} queries (Blais et al. (RANDOM 2015)). - We present a tolerant test for functions f ⁣:[n]d{0,1}f\colon[n]^d\to \{0,1\} with complexity independent of nn, which makes progress on a problem left open by Berman et al. (STOC 2014). Our techniques exploit the testing-by-learning paradigm, use novel applications of Fourier analysis on the grid [n]d[n]^d, and draw connections to distribution testing techniques.

Keywords

Cite

@article{arxiv.1609.00265,
  title  = {Testing $k$-Monotonicity},
  author = {Clément L. Canonne and Elena Grigorescu and Siyao Guo and Akash Kumar and Karl Wimmer},
  journal= {arXiv preprint arXiv:1609.00265},
  year   = {2016}
}
R2 v1 2026-06-22T15:37:44.552Z