Amortized communication complexity of distributions

Jérémie Roland, Mario Szegedy

Research output: Chapter in Book/Report/Conference proceedingConference contribution

3 Scopus citations


Consider the following general communication problem: Alice and Bob have to simulate a probabilistic function p, that with every associates a probability distribution on . The two parties, upon receiving inputs x and y, need to output , in such a manner that the (a,b) pair is distributed according to p(x,y). They share randomness, and have access to a channel that allows two-way communication. Our main focus is an instance of the above problem coming from the well known EPR experiment in quantum physics. In this paper, we are concerned with the amount of communication required to simulate the EPR experiment when it is repeated in parallel a large number of times, giving rise to a notion of amortized communication complexity. In the 3-dimensional case, Toner and Bacon showed that this problem could be solved using on average 0.85 bits of communication per repetition [1]. We show that their approach cannot go below 0.414 bits, and we give a fundamentally different technique, relying on the reverse Shannon theorem, which allows us to reduce the amortized communication to 0.28 bits for dimension 3, and 0.410 bits for arbitrary dimension. We also give a lower bound of 0.13 bits for this problem (valid for one-way protocols), and conjecture that this could be improved to match the upper bounds. In our investigation we find interesting connections to a number of different problems in communication complexity, in particular to [2]. The results contained herein are entirely classical and no knowledge of the quantum phenomenon is assumed.

Original languageEnglish (US)
Title of host publicationAutomata, Languages and Programming - 36th International Colloquium, ICALP 2009, Proceedings
Number of pages12
EditionPART 1
StatePublished - 2009
Event36th International Colloquium on Automata, Languages and Programming, ICALP 2009 - Rhodes, Greece
Duration: Jul 5 2009Jul 12 2009

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume5555 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Other36th International Colloquium on Automata, Languages and Programming, ICALP 2009

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • Computer Science(all)


Dive into the research topics of 'Amortized communication complexity of distributions'. Together they form a unique fingerprint.

Cite this