English

On Slepian--Wolf Theorem with Interaction

Computational Complexity 2015-09-09 v2

Abstract

In this paper we study interactive "one-shot" analogues of the classical Slepian-Wolf theorem. Alice receives a value of a random variable XX, Bob receives a value of another random variable YY that is jointly distributed with XX. Alice's goal is to transmit XX to Bob (with some error probability ε\varepsilon). Instead of one-way transmission, which is studied in the classical coding theory, we allow them to interact. They may also use shared randomness. We show, that Alice can transmit XX to Bob in expected H(XY)+2H(XY)+O(log2(1ε))H(X|Y) + 2\sqrt{H(X|Y)} + O(\log_2\left(\frac{1}{\varepsilon}\right)) number of bits. Moreover, we show that every one-round protocol π\pi with information complexity II can be compressed to the (many-round) protocol with expected communication about I+2II + 2\sqrt{I} bits. This improves a result by Braverman and Rao \cite{braverman2011information}, where they had 5I5\sqrt{I}. Further, we show how to solve this problem (transmitting XX) using 3H(XY)+O(log2(1ε))3H(X|Y) + O(\log_2\left(\frac{1}{\varepsilon}\right)) bits and 44 rounds on average. This improves a result of~\cite{brody2013towards}, where they had 4H(XY)+O(log1/ε)4H(X|Y) + O(\log1/\varepsilon) bits and 10 rounds on average. In the end of the paper we discuss how many bits Alice and Bob may need to communicate on average besides H(XY)H(X|Y). The main question is whether the upper bounds mentioned above are tight. We provide an example of (X,Y)(X, Y), such that transmission of XX from Alice to Bob with error probability ε\varepsilon requires H(XY)+Ω(log2(1ε))H(X|Y) + \Omega\left(\log_2\left(\frac{1}{\varepsilon}\right)\right) bits on average.

Cite

@article{arxiv.1506.00617,
  title  = {On Slepian--Wolf Theorem with Interaction},
  author = {Alexander Kozachinskiy},
  journal= {arXiv preprint arXiv:1506.00617},
  year   = {2015}
}
R2 v1 2026-06-22T09:45:14.313Z