Newman’s theorem via Carathéodory
Abstract
We give a streamlined short proof of Newman’s theorem in communication complexity by applying the classical and the approximate Carathéodory’s theorems.
Newman’s theorem is a fundamental result in communication complexity. It says that using public randomness can only gain a logarithmic advantage in communication cost comparing to private randomness. The standard proof [5, 8, 7] uses the probabilistic method and the Chernoff bound. Here, we apply Carathéodory’s theorems directly. Our proof is not too different from the standard proof, as the approximate Carathéodory is proved via a similar method. Though, the alternative proof seems neater. We note that different versions of approximate Carathéodory’s theorems and their various applications in computer science have been discussed in [1, 3].
Let denote the convex hull of a set . Let denote the usual norm. Let and denote the communication cost using private and public randomness, respectively.
Proposition 1 (Carathéodory,[2, 4]).
Let . Every can be written as a convex combination of at most points from .
Proposition 2 (Approximate Carathéodory,[1]).
Let satisfy . Let . For every , there exists , such that , and is a convex combination of at most many points from .
Theorem 1 (Newman, [6]).
Let be a Boolean function. Let . Then, .
Proof.
We view for . By definition of , there exists a set of deterministic protocols, denoted by , where is a Boolean fucntion and , satisfying
(1) |
for some where is a probabilistic average of ’s, hence, . By Proposition 1, there exists a subset of size such that . Because every is a Boolean-valued function, we have . By Proposition 2, there exists another point that can be written as a convex combination of at most points from , and
(2) |
WLOG, let , where and . Now, Alice and Bob compute using private randomness, as follows: Alice privately samples a random number according to the distribution , she sends to Bob using at most bits. Then, they run the deterministic protocol . By (1) and (2), . This gives a private randomness protocol with the desired error and cost. ∎
References
- [1] Siddharth Barman. Approximating nash equilibria and dense bipartite subgraphs via an approximate version of caratheodory’s theorem. In Proceedings of the forty-seventh annual ACM Symposium on Theory of computing, pages 361–369, 2015.
- [2] Constantin Carathéodory. Über den variabilitätsbereich der koeffizienten von potenzreihen, die gegebene werte nicht annehmen. Mathematische Annalen, 64(1):95–115, 1907.
- [3] Cyrille W Combettes and Sebastian Pokutta. Revisiting the approximate carathéodory problem via the frank-wolfe algorithm. Mathematical Programming, 197(1):191–214, 2023.
- [4] Jesús De Loera, Xavier Goaoc, Frédéric Meunier, and Nabil Mustafa. The discrete yet ubiquitous theorems of carathéodory, helly, sperner, tucker, and tverberg. Bulletin of the American Mathematical Society, 56(3):415–511, 2019.
- [5] Eyal Kushilevitz and Noam Nisan. Communication Complexity. Cambridge University Press, 1997.
- [6] Ilan Newman. Private vs. common random bits in communication complexity. Information processing letters, 39(2):67–71, 1991.
- [7] Anup Rao and Amir Yehudayoff. Communication Complexity: and Applications. Cambridge University Press, 2020.
- [8] Tim Roughgarden et al. Communication complexity (for algorithm gesigners). Foundations and Trends® in Theoretical Computer Science, 11(3–4):217–404, 2016.